machine learning

Most people have heard the terms “data science” and “AI” at least once in their lives. Indeed, both of these are extremely important in the modern world, as they are technologies that help us run quite a few of our industries. 

But even though data science and Artificial Intelligence are somewhat related to one another, they are still very different. There are things they have in common, which is why they are often used together, but it is crucial to understand their differences as well.

In this blog, we will explore the answers to data science vs AI vs machine learning, hoping to find the right demand for the advancing digital world.

What is Data Science? 

As the name suggests, data science is a field that involves studying and processing large quantities of data using a variety of technologies and techniques to detect patterns, make conclusions about the data, and aid in the decision-making process. Essentially, it is an intersection of statistics and computer science largely used in business and different industries.

The standard data science lifecycle includes capturing data and then maintaining, processing, and analyzing it before finally communicating conclusions about it through reporting. This makes data science extremely important for analysis, prediction, decision-making, problem-solving, and many other purposes. 

What is Artificial Intelligence? 

Artificial Intelligence is the field that involves the simulation of human intelligence and the processes within it by machines and computer systems. Today, it is used in a wide variety of industries and allows our society to function as it currently does by using different AI-based technologies. 

Some of the most common examples in action include machine learning, speech recognition, and search engine algorithms. While AI technologies are rapidly developing, there is still a lot of room for their growth and improvement.

For instance, there is no powerful enough content generation tool that can write texts that are as good as those written by humans. Therefore, it is always preferred to hire an experienced writer to maintain the quality of work.  

What is Machine Learning? 

As mentioned above, machine learning is a type of AI-based technology that uses data to “learn” and improve specific tasks that a machine or system is programmed to perform. Though machine learning is seen as a part of the greater field of AI, its use of data puts it firmly at the intersection of data science and AI.

Similarities Between Data Science and AI 

By far the most important point of connection between data science and Artificial Intelligence is data. Without data, neither of the two fields would exist, and the technologies within them would not be used so widely in all kinds of industries.

In many cases, data scientists and AI specialists work together to create new technologies, improve old ones, and find better ways to handle data. 

As explained earlier, there is a lot of room for improvement when it comes to AI technologies. The same can be somewhat said about data science. That’s one of the reasons businesses still hire professionals to accomplish certain tasks, like custom writing requirements, design requirements, and other administrative work.

Differences Between Data Science and AI

There are quite a few differences between both. These include:

Purpose – It aims to analyze data to make conclusions, predictions, and decisions. Artificial Intelligence aims to enable computers and programs to perform complex processes in a similar way to how humans do. 

Scope – This includes a variety of data-related operations such as data mining, cleansing, reporting, etc. It primarily focuses on machine learning, but there are other technologies involved too such as robotics, neural networks, etc. 

Application – Both are used in almost every aspect of our lives, but while data science is predominantly present in business, marketing, and advertising, AI is used in automation, transport, manufacturing, and healthcare. 

Examples of Data Science and Artificial Intelligence in Use 

To give you an even better idea of what data science and Artificial Intelligence are used for, here are some of the most interesting examples of their application in practice: 

• Analytics – Analyze customers to better understand the target audience and offer the kind of product or service that the audience is looking for. 
• Monitoring – Monitor the social media activity of specific types of users and analyze their behavior. 
• Prediction – Analyze the market and predict demand for specific products or services in the nearest future. 
• Recommendation – Recommend products and services to customers based on their customer profiles, buying behavior, etc. 

• Forecasting – Predict the weather based on a variety of factors and then use these predictions for better decision-making in the agricultural sector. 
• Communication – Provide high-quality customer service and support with the help of chatbots. 
• Automation – Automate processes in all kinds of industries, from retail and manufacturing to email marketing and pop-up on-site optimization. 
• Diagnosing – Identify and predict diseases, give correct diagnoses, and personalize healthcare recommendations. 
• Transportation – Use self-driving cars to get where you need to go. Use self-navigating maps to travel. 

• Assistance – Get assistance from smart voice assistants that can schedule appointments, search for information online, make calls, play music, and more. 
• Filtering – Identify spam emails and automatically get them filtered into the spam folder. 
• Cleaning – Get your home cleaned by a smart vacuum cleaner that moves around on its own and cleans the floor for you. 
• Editing – Check texts for plagiarism, proofread, and edit them by detecting grammatical, spelling, punctuation, and other linguistic mistakes. 

It is not always easy to tell which of these examples is about data science and which one is about Artificial Intelligence because many of these applications use both of them. This way, it becomes even clearer just how much overlap there is between these two fields and the technologies that come from them. 

Data Science vs AI vs ML: What is Your Choice?

At the end of the day, data science and AI remain some of the most important technologies in our society and will likely help us invent more things and progress further. As a regular citizen, understanding the similarities and differences between the two will help you better understand how data science and Artificial Intelligence are used in almost all spheres of our lives. 

In this blog, we will discuss the top 8 Machine Learning algorithms that will help you to receive and analyze input data to predict output values within an acceptable range

1. Linear Regression 

Linear regression is a simple machine learning model and chances are you are already aware of it! Do you remember plotting the line y=mx+c in your introductory algebra class? This is an equation of a straight line where m is its gradient and c is the point where the line crosses the y-axis. Using this equation, you’re able to estimate the value of y for any given value of x. Similarly, linear regression involves estimating the relationship between independent variables (x) and a dependent variable(y).  

2. Logistic Regression 

Just like linear regression, logistic regression is a machine learning model used to determine the relationship between a dependent variable and one or more independent variables. However, this model is used for classification analysis. This is because logistic regression predicts the probability of an event occurring. For a probability greater than 0.5, a value of 1 is assigned, and for less than that 0. For example, you can use logistic regression to predict whether a student will pass (1) an exam, or they will fail (0). 

3. Decision Trees 

Decision tree is a supervised machine learning model that repeatedly splits the data based on a question corresponding to the features. The model learns the best way to reduce randomness and drafts a decision tree that can be used to predict the category of an item based on answering a selection of questions. For example, in the case of whether it will rain today or not, the questions can be whether it is sunny, did it rain yesterday, whether it is windy, and so on.  

4. Random Forest 

Random Forest is a machine learning algorithm that works similarly to a decision tree. The difference is that random forest uses multiple decision trees to make a prediction and hence decreases overfitting. The process of majority voting is carried out and the class selected by most trees is assigned to an item. For example, if two trees predict it to be 0, and one tree predicts it to be 1, then the class of 0 will be assigned to the item.  

5. K-Nearest Neighbor 

K-Nearest Neighbor is another simple machine learning algorithm that classifies new cases based on the category/class of the data points nearest to the new data point. That is, if most neighbors of an unknown item belong to class 1, then we assign class 1 to this unknown item. The number of neighbors to take into consideration is the value K assigned. If k=10, we will look at the 10 nearest neighbors of this item. The nearest neighbors are determined by measuring the distance using distance measures such as Euclidean distance, and the nearest are those that have the shortest distance. 

6. Support Vector Machine 

Support vector machines by dividing the data points using a hyperplane which is a straight line. The points donated by the blue diamond form one class on the left side of the plane and the points donated by the green circle represent another class on the right side of the plane. If we want to predict the class of a new point, we can simply determine it by whether it lies on the left or right side of the hyperplane and where it is within the margin. 

7. K-Means clustering 

K-means clustering is an unsupervised machine learning algorithm. That means it is used to work with data points whose class is not already known. We can use the clustering algorithm to group similar items into clusters. The number of clusters is determined by the value of K assigned. For example, you assign K=3. Three clusters are selected at random, and we adjust them until they are highly distinct from one another. Distinct clusters will have points similar to each other but these points will be distinct from points in another cluster.

8. Naïve Bayes

Naïve Bayes is a probabilistic machine learning model based on the Bayes theorem that assumes that all the features are independent of one another. Conditional probability refers to the probability of an outcome occurring if it is given that another event has occurred. This algorithm predicts the probability that an item belongs to a particular class and is assigned the class with the highest probability. 

Share more Machine Learning algorithms with us

Have we missed any Machine Learning algorithm that you would like to learn about? Share with us in the comments below

Learning Data Science with fun is the missing ingredient for diligent data scientists. This blog post collected the best data science jokes including statistics, artificial intelligence, and machine learning.

For Data Scientists

1. There are two kinds of data scientists. 1.) Those who can extrapolate from incomplete data.

2. Data science is 80% preparing data, and 20% complaining about preparing data.

3. There are 10 kinds of people in this world. Those who understand binary and those who don’t.

4. What’s the difference between an introverted data analyst & an extroverted one? Answer: the extrovert stares at YOUR shoes.

5. Why did the chicken cross the road? The answer is trivial and is left as an exercise for the reader.

Here’s this also for data scientists: 6 Books to Help You Learn Data Science

6. The data science motto: If at first, you don’t succeed; call it version 1.0

7. What do you get when you cross a pirate with a data scientist? Answer: Someone who specializes in Rrrr

8. A SQL query walks into a bar, walks up to two tables, and asks, “Can I join you?”

9. Why should you take a data scientist with you into the jungle? Answer: They can take care of Python problems

10. Old data analysts never die – they just get broken down by age

11. I don’t know any programming, but I still use Excel in my field!

12. Data is like people – interrogate it hard enough and it will tell you whatever you want to hear.

13. Don’t get it? We can help. Check out our in-person data science Bootcamp or online data science certificate program.

For Statisticians

14. Statistics may be dull, but it has its moments.

15. You are so mean that your standard deviation is zero.

16. How did the random variable get into the club? By showing a fake I.D.

17. Did you hear the one about the statistician? Probably….

18. Three statisticians went out hunting and came across a large deer. The first statistician fired, but missed, by a meter to the left. The second statistician fired, but also missed, by a meter to the right. The third statistician didn’t fire, but shouted in triumph, “On average we got it!”

19. Two random variables were talking in a bar. They thought they were being discreet, but I heard their chatter continuously.

20. Statisticians love whoever they spend the most time with; that’s their statistically significant other.

21. Old age is statistically good for you – very few people die past the age of 100.

22. Statistics prove offspring is an inherited trait. If your parents didn’t have kids, odds are you won’t either.

For Artificial Intelligence experts

23. Artificial intelligence is no match for natural stupidity

24. Do neural networks dream of strictly convex sheep?

25. What did one support vector say to another support vector? Answer: I feel so marginalized

Here are some of the AI memes and jokes you wouldn’t want to miss

26. AI blogs are like philosophy majors. They’re always trying to explain “deep learning.”

27. How many support vectors does it take to change a light bulb? Answer: Very few, but they must be careful not to shatter* it.

28. Parent: If all your friends jumped off a bridge, would you follow them? Machine Learning Algorithm: yes.

29. They call me Dirichlet because all my potential is latent and awaiting allocation

30. Batch algorithms: YOLO (You Only Learn Once), Online algorithms: Keep Updates and Carry On

Read up on the 10 Must-Have AI Engineering Skills

31. “This new display can recognize speech” “What?” “This nudist play can wreck a nice beach”

32. Why did the naive Bayesian suddenly feel patriotic when he heard fireworks? Answer: He assumed independence

33. Why did the programmer quit their job? Answer: Because they didn’t get arrays.

34. What do you call a program that identifies spa treatments? Facial recognition!

35. Human: What do we want!?

• Computer: Natural language processing!
• Human: When do we want it!?
• Computer: When do we want what?

36. A statistician’s wife had twins. He was delighted. He rang the minister who was also delighted. “Bring them to church on Sunday and we’ll baptize them,” said the minister. “No,” replied the statistician. “Baptize one. We’ll keep the other as a control.”

For Machine Learning Professionals

37. I have a joke about a data miner, but you probably won’t dig it. @KDnuggets:

38. I have a joke about deep learning, but I can’t explain it. Shamail Saeed, @hacklavya

39. I have a joke about deep learning, but it is shallow. Mehmet Suzen, @memosisland

40. I have a machine learning joke, but it is not performing as well on a new audience. @dbredesen

41. I have a new joke about Bayesian inference, but you’d probably like the prior more. @pauljmey

42. I have a joke about Markov models, but it’s hidden somewhere. @AmeyKUMAR1

43. I have a statistics joke, but it’s not significant. @micheleveldsman

Explore this Comprehensive Guide to Machine Learning

44. I have a geography joke, but I don’t know where it is. @olimould

45. I have an object-oriented programming joke. But it has no class. Ayin Vala

46. I have a quantum mechanics joke. It’s both funny and not funny at the same time. Philip Welch

47. I have a good Bayesian laugh that came from a prior joke. Nikhil Kumar Mishra

48. I have a Java joke, but it is too verbose! Avneesh Sharma

49. I have a regression joke, but it sounds quite mean. Gang Su

50. I have a machine-learning joke, but I cannot explain it. Andriy Burkov

Do You Have any Data Science Jokes to Share?

Share your favorite data science jokes with us in the comments below. Let’s laugh together!

Data Science Dojo is offering Apache Zeppelin for FREE on Azure Marketplace packaged with pre-installed interpreters and backends to make Machine Learning easier than ever. 

Introduction 

How cumbersome and tiring it is to install different tools to perform your desired ML tasks and then look after the integration and dependency issues. Already getting headaches? Worry not, because Data Science Dojo’s Apache Zeppelin instance fixes all of that. But before we delve further into it, let’s get to know some basics. 

What are Machine Learning Operations?  

Machine Learning is a branch of Artificial Intelligence that deals with models that produce outcomes based on some learned pre-existing data. It provides automation and reduces the workload of users. ML converges with Data Science and Engineering and that gives birth to some necessary operations to be performed to acquire the output of any task.

These operations include ETL (Extraction, Transform, Load) or ELT, drawing interactive visualizations, running queries, training and testing ML models and several other functions. 

Pro Tip: Join our 6-months instructor-led Data Science Bootcamp to master machine learning skills. 

Challenges for individuals 

 Wanting to explore and visualize your data but not knowing the methodology of the new tool is not only a red flag but also demands extraneous skills to be learnt to proceed with your job. Or you would have to switch among different environments to achieve your goal which is again – time-consuming, and needless to say time is of the essence for data scientists and engineers when they must deliver a task.

In this scenario, switching from one tool to another which you may know how to use or may not, is time and cost intensive. What if a data driven interactive environment having several interpreters ready to be worked with in one place is provided and you just select your favorite language and break the ice? 

ML Operations with Apache Zeppelin 

Apache Zeppelin is an open-source tool that equips you with a web-based notebook that can be used for data processing and querying, handling big data, training and testing models, interactive data analytics, visualization, and exploration. Vibrant designs and pictures generated can save time for users in the identification of key patterns in data and ultimately accelerates the decision-making processes.

It contains different pre-installed interpreters but also allows you to plug in your own various language backends for desirability. Apache Zeppelin supports many data sources which allow you to synthesize your data to visualize into interactive plots and charts. You can also create dynamic forms in your notebook and can share your notebook with collaborators.              

(Picture Courtesy: https://zeppelin.apache.org/ ) 

Key features 

• Zeppelin delivers an optimized and interactive UI that enhances the plots, charts, and other diagrams. You can also create dynamic forms in your notebook along with other markdowns to fancify your note 
• It’s open-source and allows vendors to make Zeppelin highly customized according to use-case requirements that vary from industry to industry 

• The choice to select a learned backend from a variety of pre-installed ones or the feasibility to add your own customizable language adds to the user-friendliness, flexibility, and adaptability 
• It supports Big Data databases like Hive and Spark. It also provides support for web sockets so you can share your web page by echoing the output of the browser and creating live reports 
• Zeppelin provides an in-build job manager who keeps track of the condition or status of various notebooks 

What Data Science Dojo has for you 

Our Zeppelin instance serves as a web-accessible programming environment with miscellaneous pre-installed interpreters. In our service users can switch between different interpreters like processing data with python and then visualizing it by querying with SQL. The pre-installed backends provide the feasibility to perform the task using your accustomed language instead of learning a new tool. 

• A web-accessible Zeppelin environment 
• Several pre-installed language-backends/interpreters 
• Various tutorial notebooks containing codes for understandability 

• A Job manager responsible for monitoring the status of the notebooks 
• A Notebook Repos feature to manage your notebook repositories’ settings 
• Ability to import notes from JSON file or URL 
• In-build functionality to add and modify your own customized interpreters 
• Credential management service 

Our instance supports the following interpreters: 

• Alluxio 
• Angular 
• Beam 
• BigQuery 

• Cassandra 
• Elasticsearch 
• File 
• Flink 

And many others which you check by taking a quick peek here: Zeppelin on Market Place  

Conclusion 

Efficient resource requirement for processing, visualizing, and training large data was one of the areas of concern when working on traditional desktop environments. The other area of concern includes the burden of working with non-familiar backends or switching among different accustomed environments. With our Zeppelin instance, both concerns are put to rest.

When coupled with Microsoft Azure services and processing speed, it outperforms the traditional counterparts because data-intensive computations aren’t performed locally, but in the cloud. You can collaborate and share notebooks with various stakeholders within and outside the company while monitoring the status of each 

At Data Science Dojo, we deliver data science education, consulting, and technical services to increase the power of data. We are therefore adding a free Zeppelin Notebook Environment dedicated specifically for Machine Learning and Data Science operations on Azure Market Place. Don’t wait to install this offer by Data Science Dojo, your ideal companion in your journey to learn data science! 

Click on the button below to head over to the Azure Marketplace and deploy Apache Zeppelin for FREE by clicking on “Get it now”.

Note: You’ll have to sign up to Azure, for free, if you do not have an existing account.

Be it Netflix, Amazon, or another mega-giant, their success stands on the shoulders of experts, analysts are busy deploying machine learning through supervised, unsupervised, and reinforcement successfully. 

The tremendous amount of data being generated via computers, smartphones, and other technologies can be overwhelming, especially for those who do not know what to make of it. To make the best use of data researchers and programmers often leverage machine learning for an engaging user experience.

Many advanced techniques that are coming up every day for data scientists of all supervised, and unsupervised, reinforcement learning is leveraged often. In this article, we will briefly explain what supervised, unsupervised, and reinforcement learning is, how they are different, and the relevant uses of each by well-renowned companies.

Supervised learning

Supervised machine learning is used for making predictions from data. To be able to do that, we need to know what to predict, which is also known as the target variable. The datasets where the target label is known are called labeled datasets to teach algorithms that can properly categorize data or predict outcomes. Therefore, for supervised learning:

• We need to know the target value
• Targets are known in labeled datasets

Let’s look at an example: If we want to predict the prices of houses, supervised learning can help us predict that. For this, we will train the model using characteristics of the houses, such as the area (sq ft.), the number of bedrooms, amenities nearby, and other similar characteristics, but most importantly the variable that needs to be predicted – the price of the house.

A supervised machine learning algorithm can make predictions such as predicting the different prices of the house using the features mentioned earlier, predicting trends of future sales, and many more.

Sometimes this information may be easily accessible while other times, it may prove to be costly, unavailable, or difficult to obtain, which is one of the main drawbacks of supervised learning.

Saniye Alabeyi, Senior Director Analyst at Garnet calls Supervised learning the backbone of today’s economy, stating:

“Through 2022, supervised learning will remain the type of ML utilized most by enterprise IT leaders” (Source).

Types of problems:

Supervised learning deals with two distinct kinds of problems:

1. Classification problems
2. Regression problems

Classification problem: In the case of classification problems, examples are classified into one or more classes/ categories.

For example, if we are trying to predict that a student will pass or fail based on their past profile, the prediction output will be “pass/fail.” Classification problems are often resolved using algorithms such as Naïve Bayes, Support Vector Machines, Logistic Regression, and many others.

Regression problem: A problem in which the output variable is either a real or continuous value, s is defined as a regression problem. Bringing back the student example, if we are trying to predict that a student will pass or fail based on their past profuse, the prediction output will be numeric, such as “68%” likely to score.

Predicting the prices of houses in an area is an example of a regression problem and can be solved using algorithms such as linear regression, non-linear regression, Bayesian linear regression, and many others.

Here’s a comprehensive guide to Machine Learning Model Deployment

Why Amazon, Netflix, and YouTube are great fans of supervised learning?

Recommender systems are a notable example of supervised learning. E-commerce companies such as Amazon, streaming sites like Netflix, and social media platforms such as TikTok, Instagram, and even YouTube among many others make use of recommender systems to make appropriate recommendations to their target audience.

Unsupervised learning

Imagine receiving swathes of data with no obvious pattern in it. A dataset with no labels or target values cannot come up with an answer to what to predict. Does that mean the data is all waste? Nope! The dataset likely has many hidden patterns in it.

Unsupervised learning studies the underlying patterns and predicts the output. In simple terms, in unsupervised learning, the model is only provided with the data in which it looks for hidden or underlying patterns.

Unsupervised learning is most helpful for projects where individuals are unsure of what they are looking for in data. It is used to search for unknown similarities and differences in data to create corresponding groups.

An application of unsupervised learning is the categorization of users based on their social media activities.

Commonly used unsupervised machine learning algorithms include K-means clustering, neural networks, principal component analysis, hierarchical clustering, and many more.

Reinforcement learning

Another type of machine learning is reinforcement learning.

In reinforcement learning, algorithms learn in an environment on their own. The field has gained quite some popularity over the years and has produced a variety of learning algorithms.

Reinforcement learning is neither supervised nor unsupervised as it does not require labeled data or a training set. It relies on the ability to monitor the response to the actions of the learning agent.

Most used in gaming, robotics, and many other fields, reinforcement learning makes use of a learning agent. A start state and an end state are involved. For the learning agent to reach the final or end stage, different paths may be involved.

• An agent may also try to manipulate its environment and may travel from one state to another
• On success, the agent is rewarded but does not receive any reward or appreciation for failure
• Amazon has robots picking and moving goods in warehouses because of reinforcement learning

Also learn about Retrieval Augmented Generation

Numerous IT companies including Google, IBM, Sony, Microsoft, and many others have established research centers focused on projects related to reinforcement learning.

Social media platforms like Facebook have also started implementing reinforcement learning models that can consider different inputs such as languages, integrate real-world variables such as fairness, privacy, and security, and more to mimic human behavior and interactions. (Source)

Amazon also employs reinforcement learning to teach robots in its warehouses and factories how to pick up and move goods.

Comparison between supervised, unsupervised, and reinforcement learning

Caption: Differences between supervised, unsupervised, and reinforcement learning algorithms

Which is the better Machine Learning technique?

We learned about the three main members of the machine learning family essential for deep learning. Other kinds of learning are also available such as semi-supervised learning, or self-supervised learning.

Supervised, unsupervised, and reinforcement learning, are all used for different to complete diverse kinds of tasks. No single algorithm exists that can solve every problem, as problems of different natures require different approaches to resolve them.

Despite the many differences between the three types of learning, all of these can be used to build efficient and high-value machine learning and Artificial Intelligence applications. All techniques are used in different areas of research and development to help solve complex tasks and resolve challenges.

If you would like to learn more about data science, machine learning, and artificial intelligence, visit the Data Science Dojo blog.

Written by Alyshai Nadeem

Confused about which machine learning conferences you should attend? Here are our top 10 picks for the remaining months of 2022.

For aspiring data scientists, machine learners, and researchers, conferences are a great way to network, highlight their own work, and learn from others. This article highlights the top 10 machine learning conferences that you should attend if you are in Asia or are planning to visit soon.

1. ACAIT 2022: The 6th Asian Conference on Artificial Intelligence Technology – Changzhou, China

Taking place in the southern Jiangsu province of China, on the 4th of November, the ACAIT is a joint endeavor of the Institute of Electrical and Electronics Engineers (IEEE), Chinese Association for Artificial Intelligence (CAAI), and Changzhou Institute of Technology (CIT).

The conference invites significant and original research work from the world of artificial intelligence. The main aim of the conference is to provide an international forum for researchers to share their ideas and achievements in the field of artificial intelligence.

The conference covers all major topics from AI-related brain and cognitive sciences to machine Cognition and Pattern Recognition, Big data and knowledge engineering, Robotics, swarm intelligence, and even the Internet of Things.

Further details regarding the conference can be found here.

2. 4th Asian Conference on Machine Learning (ACML 2022) – Hyderabad, India

Taking place between 12th to 14th December in Hyderabad, India, the ACML abides by the post-pandemic laws and will be conducted virtually, as well as allow in-person interactions.

Focusing on theoretical and practical aspects of machine learning, the conference encourages researchers from around the globe to join and be a part of the conversation.

The conference will cover general machine learning topics such as supervised learning and reinforcement learning, and even dive deeper into Deep Learning, Probabilistic Methods, theoretical frameworks, and much more.

Further details regarding the conference can be found here.

3. The 29th International Conference on Computational Linguistics – Gyeongju, Republic of Korea

One of the most popular conferences on natural language processing and computational linguistics, COLING is expected to be held on October 12-17, 2022, in Gyeongju, South Korea.

The conference has been held every year since 1965. Participants from both top-ranked research centers and emerging countries attend this conference as it provides equal opportunities to researchers from educational institutes and academia, as well as from the corporate sector.

COLING focuses on all aspects of natural language processing and computation.

Not only is this one of the most prestigious conferences on NLP and computational linguistics, but it is also heavily sponsored by names such as LG Electronics, Hyundai, Google, and Apple, among many others.

Further details regarding the conference can be found here.

4. IROS 2022: International Conference on Intelligent Robots and Systems – Kyoto, Japan

One of the flagship conferences of the robotics community, IROS is one of the world’s oldest forums for the global robotics community to explore intelligent robots and systems. Held every year in Kyoto, Japan since 1987, the conference will be held on 23-27 October.

Not only does the conference feature numerous research works from various international authors, but the conference also includes workshops and training, as well as multiple guest lectures by professionals in academia and industry.

Further details regarding the conference can be found here.

5. ACCV 2022: The 16th Asian Conference on Computer Vision

The Asian Conference on Computer Vision (AACV) 2022 focuses on computer vision and pattern recognition and will be held on 4-8 December in Macau, China.

The biennial international conference is sponsored by the Asian Federation of Computer Vision and provides like-minded individuals an opportunity to discuss the latest problems, solutions, and technologies in the field of computer vision and other similar areas.

The conference proceedings are published by Springer as Lecture Notes. Moreover, the award-winning papers are invited for publication in a special issue of the International Journal of Computer Vision (IJCV).

More details on the conference can be found here.

6. The 29th International Conference on Neural Information Processing (ICONIP 2022), New Delhi, India

One of the leading international conferences in the fields of pattern recognition, neuroscience, intelligent control, information security, and brain-machine interface, the ICONIP will be held in New Delhi, India on 22nd -26th November 2022.

It is the annual flagship conference organized by the Asia Pacific Neural Network Society (APNNS), which strives towards bridging the gap between educational institutions and industry.

The conference provides an international forum for anyone working in neuroscience, neural networks, deep learning, and other similar fields.

The conference is divided into four categories: Theory and Algorithms, Computational and Cognitive Neurosciences, Human-Centered Computing, and other machine learning applications.

Further details on the conference can be found here.

7. The 19th Pacific Rim International Conference on Artificial Intelligence (PRICAI) – Shanghai, China

A biennial international conference, the PRICAI focuses on AI theories, technologies, and their applications in areas of social and economic importance, specifically focusing on countries in the Pacific Rim. Held since 1990, PRICAI will take place on 10-13th November, in the financial hub of China – Shanghai.

The conference focuses on all things related to AI, machine learning, data mining, robotics, computer vision, and much more.

Further information regarding the conference can be found here.

8. The 4th International Conference on Data-driven Optimization of Complex Systems (DOCS2022) – Chengdu, China

Focused on data-driven optimization, learning and control, and their applications to complex systems, DOCS 2022 will be held 23-25th September, Chengdu, Sichuan, China.

The conference focuses on topics ranging from data-driven machine learning, optimization, decision-making, analysis, and application.

Further details on the conference can be found here.

9. The 9th IEEE International Conference on Data Science and Advanced Analytics (DSAA) – Shenzhen, China

Widely recognized as a dedicated flagship annual conference, the International Conference on Data Science and Advanced Analytics (DSAA) will be held in Shenzhen, China on the 13th –16th of October 2022.

The conference not only focuses on computing and information/intelligence sciences but also considers their relationship with statistics, and the crossover of data science and analytics.

An interesting aspect of this conference is that it is a dual-track conference with both a research track and an application track. Further details regarding these different tracks can be found here.

More details on the conference can be found here.

10. The 5th International Conference on Intelligent Autonomous Systems (ICoIAS 2022) – Dalian, China

The ICoIAS conference focuses on intelligent autonomous systems that play a significant role in multiple control and engineering applications.

The conference will be held on 23-25 September at the Dalian Maritime University, Dalian, China, in collaboration with Tianjin University, the IEEE Computational Intelligence Society, and The Institution of Engineers, Singapore.

The conference focuses on distinct aspects of intelligent autonomous systems. Moreover, IEEE fellows from all over the world are expected to attend the conference as guest speakers.

For further information regarding the conference, click here.

Was this list helpful? Let us know in the comments below. If you would like to find similar conferences in a different area, click here.

If you are interested in learning more about machine learning and data science, click here.

Written by Alyshai Nadeem

Introduction

Data Scientists use machine learning techniques to make predictions under a variety of scenarios. Machine learning can be used to predict whether a borrower will default on his mortgage or not, or what might be the median house value in a given zip code area. Depending upon whether the prediction is being made for a quantitative variable or a qualitative variable, a predictive model can be categorized as a regression model (e.g. predicting median house values) or a classification (e.g. predicting loan defaults) model.

Decision trees happen to be one of the simplest and easiest classification models to explain and, as many argue, closely resemble human decision-making.

This tutorial has been developed to help you revisit and master the fundamentals of decision tree classification models which are expanded on in Data Science Dojo’s data science bootcamp and online data science certificate program. Our key focus will be to discuss the:

1. Fundamental concepts on data-partitioning, recursive binary splitting, nodes, etc.
2. Data exploration and data preparation for building classification models
3. Performance metrics for decision tree models – Gini Index, Entropy, and Classification Error.

The content builds your classification model knowledge and skills in an intuitive and gradual manner.

The scenario

You are a Data Scientist working at the Centers for Disease Control (CDC) Division for Heart Disease and Stroke Prevention. Your division has recently completed a research study to collect health examination data among 303 patients who presented with chest pain and might have been suffering from heart disease.

The Chief Data Scientist of your division has asked you to analyze this data and build a predictive model that can accurately predict patients’ heart disease status, identifying the most important predictors of heart failure. Once your predictive model is ready, you will make a presentation to the doctors working at the health facilities where the research was conducted.

The data set has 14 attributes, including patients’ age, gender, blood pressure, cholesterol level, and heart disease status, indicating whether the diagnosed patient was found to have heart disease or not. You have already learned that to predict quantitative attributes such as “blood pressure” or “cholesterol level”, regression models are used, but to predict a qualitative attribute such as the “status of heart disease,”  classification models are used.

Classification models can be built using different techniques such as Logistic Regression, Discriminant Analysis, K-Nearest Neighbors (KNN), Decision Trees, etc. Decision Trees are very easy to explain and can easily handle qualitative predictors without the need to create dummy variables.

Although decision trees generally do not have the same level of predictive accuracy as the K-Nearest Neighbor or Discriminant Analysis techniques, They serve as building blocks for other sophisticated classification techniques such as “Random Forest” etc. which makes mastering Decision Trees, necessary!

We will now build decision trees to predict the status of heart disease i.e. to predict whether the patient has heart disease or not, and we will learn and explore the following topics along the way:

• Data preparation for decision tree models
• Classification trees using “rpart” package
• Pruning the decision trees
• Evaluating decision tree models

## You will need following libraries for this exercise 

library(dplyr) 
library(tidyverse)
library(ggplot2)
library(rpart)
library(rpart.plot)
library(rattle)
library(RColorBrewer)

## Following code will help you suppress the messages and warnings during package loading      
options(warn = -1) 

The data

You will be working with the Heart Disease Data Set which is available at UC Irvine’s Machine Learning Repository. You are encouraged to visit the repository and go through the data description. As you will find, the data folder has multiple data files available. You will use the processed.cleveland.data.

Let’s read the datafile into a data frame “cardio”

## Reading the data into "cardio" data frame
cardio <- read.csv("processed.cleveland.data", header = FALSE, na.strings = '?')            

## Let's look at the first few rows in the cardio data frame  
head(cardio)

As you can see, this data frame doesn’t have column names. However, we can refer to the data dictionary, given below, and add the column names:

The following code chunk will add column names to your data frame:

## Adding column names to dataframe 
names(cardio) <- c( "age", "sex", "cp", "trestbps", "chol","fbs", "restecg", 
                           "thalach","exang", "oldpeak","slope", "ca", "thal", "status")

You are going to build a decision tree model to predict values under variable #14 status, the “angiographic disease status” which labels or classifies each patient as “having heart disease” or “not having heart disease.

Intuitively, we expect some of these other 13 variables to help us predict the values under status. In other words, we expect variables #1 to #13, to segment the patients or create partitions in the cardio data frame in a manner that any given partition (or segment) thus created either has patients as “having heart disease” or “not having heart disease.

Data preparation for decision trees

It is time to get familiar with the data. Let’s begin with data types.

## We will use str() function  
str(cardio)

'data.frame':	303 obs. of  14 variables:
 $ age      : num  63 67 67 37 41 56 62 57 63 53 ...
 $ sex      : num  1 1 1 1 0 1 0 0 1 1 ...
 $ cp       : num  1 4 4 3 2 2 4 4 4 4 ...
 $ trestbps : num  145 160 120 130 130 120 140 120 130 140 ...
 $ chol     : num  233 286 229 250 204 236 268 354 254 203 ...
 $ fbs      : num  1 0 0 0 0 0 0 0 0 1 ...
 $ restecg  : num  2 2 2 0 2 0 2 0 2 2 ...
 $ thalach  : num  150 108 129 187 172 178 160 163 147 155 ...
 $ exang    : num  0 1 1 0 0 0 0 1 0 1 ...
 $ oldpeak  : num  2.3 1.5 2.6 3.5 1.4 0.8 3.6 0.6 1.4 3.1 ...
 $ slope    : num  3 2 2 3 1 1 3 1 2 3 ...
 $ ca       : num  0 3 2 0 0 0 2 0 1 0 ...
 $ thal     : num  6 3 7 3 3 3 3 3 7 7 ...
 $ status   : int  0 2 1 0 0 0 3 0 2 1 ...

As you can see, some qualitative variables in our data frame are included as quantitative variables

• status is declared as $$ which makes it a quantitative variable but we know disease status must be qualitative
• You can see that sex, cp, fbs, restecg, exang,  slope, ca, and thal too
  must be qualitative

The next code-chunk will convert and correct the datatypes:

## We can use lapply to convert data types across multiple columns  
cardio[c("sex", "cp", "fbs","restecg", "exang", 
                     "slope", "ca", "thal", "status")] <- lapply(cardio[c("sex", "cp", "fbs","restecg",
                                                                         "exang", "slope", "ca", "thal", "status")], factor)
## You can verify the data frame 
str(cardio)

'data.frame':	303 obs. of  14 variables:
 $ age     : num  63 67 67 37 41 56 62 57 63 53 ...
 $ sex     : Factor w/ 2 levels "0","1": 2 2 2 2 1 2 1 1 2 2 ...
 $ cp      : Factor w/ 4 levels "1","2","3","4": 1 4 4 3 2 2 4 4 4 4 ...
 $ trestbps: num  145 160 120 130 130 120 140 120 130 140 ...
 $ chol    : num  233 286 229 250 204 236 268 354 254 203 ...
 $ fbs     : Factor w/ 2 levels "0","1": 2 1 1 1 1 1 1 1 1 2 ...
 $ restecg : Factor w/ 3 levels "0","1","2": 3 3 3 1 3 1 3 1 3 3 ...
 $ thalach : num  150 108 129 187 172 178 160 163 147 155 ...
 $ exang   : Factor w/ 2 levels "0","1": 1 2 2 1 1 1 1 2 1 2 ...
 $ oldpeak : num  2.3 1.5 2.6 3.5 1.4 0.8 3.6 0.6 1.4 3.1 ...
 $ slope   : Factor w/ 3 levels "1","2","3": 3 2 2 3 1 1 3 1 2 3 ...
 $ ca      : Factor w/ 4 levels "0","1","2","3": 1 4 3 1 1 1 3 1 2 1 ...
 $ thal    : Factor w/ 3 levels "3","6","7": 2 1 3 1 1 1 1 1 3 3 ...
 $ status  : Factor w/ 5 levels "0","1","2","3",..: 1 3 2 1 1 1 4 1 3 2 ...

Also, note that status has 5 different values viz. 0, 1, 2, 3, 4. While status = 0, indicates no heart disease, all other values under status indicate a heart disease. In this exercise, you are building a decision tree model to classify each patient as “normal”(not having heart disease) or “abnormal” (having heart disease)”.

Therefore, you can merge status = 1, 2, 3, and 4 into a single-level status = “1”. This way you will convert status into a  Binary or Dichotomous variable having only two values status = “0” (normal) and status = “1” (abnormal)

Let’s do that!

##  We will use the 'forcats' package included in the s'tidyverse' package
##  The function to be used will be fct_collpase 
cardio$status <- fct_collapse(cardio$status, "1" = c("1","2", "3", "4"))  


## Let's also change the labels under the "status" from (0,1) to (normal, abnormal)  
levels(cardio$status) <- c("normal", "abnormal")  

## levels under sex can also be changed to (female, male)   
## We can change level names in other categorical variables as well but we are not doing that  
levels(cardio$sex) <- c("female", "male")  

So, you have corrected the data types. What’s next?

How about getting a summary of all the variables in the data?

## Overall summary of all the columns 
summary(cardio)

      age            sex      cp         trestbps          chol       fbs    
 Min.   :29.00   female: 97   1: 23   Min.   : 94.0   Min.   :126.0   0:258  
 1st Qu.:48.00   male  :206   2: 50   1st Qu.:120.0   1st Qu.:211.0   1: 45  
 Median :56.00                3: 86   Median :130.0   Median :241.0          
 Mean   :54.44                4:144   Mean   :131.7   Mean   :246.7          
 3rd Qu.:61.00                        3rd Qu.:140.0   3rd Qu.:275.0          
 Max.   :77.00                        Max.   :200.0   Max.   :564.0

 restecg    thalach      exang      oldpeak     slope      ca        thal    
 0:151   Min.   : 71.0   0:204   Min.   :0.00   1:142   0   :176   3   :166  
 1:  4   1st Qu.:133.5   1: 99   1st Qu.:0.00   2:140   1   : 65   6   : 18  
 2:148   Median :153.0           Median :0.80   3: 21   2   : 38   7   :117  
         Mean   :149.6           Mean   :1.04           3   : 20   NA's:  2  
         3rd Qu.:166.0           3rd Qu.:1.60           NA's:  4             
         Max.   :202.0           Max.   :6.20                                

       status   
 normal  :164  
 abnormal:139  

Did you notice the missing values (NAs) under the ca and thal columns? With the following code, you can count the missing values across all the columns in your data frame.

# Counting the missing values in the datframe 
sum(is.na(cardio))

6

Only 6 missing values across 303 rows which is approximately 2%. That seems to be a very low proportion of missing values. What do you want to do with these missing values, before you start building your decision tree model?

• Option 1: discard the missing values before training.
• Option 2: rely on the machine learning algorithm to deal with missing values during the model training.
• Option 3: impute missing values before training.

For most learning methods, Option 3 the imputation approach is necessary. The simplest approach is to impute the missing values by the mean or median of the non-missing values for the given feature.

The choice of Option 2 depends on the learning algorithm. Learning algorithms such as CART and rpart simply ignore missing values when determining the quality of a split. To determine, whether a case with a missing value for the best split is to be sent left or right, the algorithm uses surrogate splits. You may want to read more on this here.

However, if the relative amount of missing data is small, you can go for Option 1 and discard the missing values as long as it doesn’t lead to or further alleviate the class imbalance which is briefly discussed in the following section.

As for your data set, you are safe to delete missing value cases. The following code-chunk does that for you.

## Removing missing values  
cardio <- na.omit(cardio)

Data exploration

Status is the variable that you want to predict with your model. As we have discussed earlier, other variables in the cardio dataset should help you predict status.

For example, amongst patients with heart disease, you might expect the average value of Cholesterol levels (chol), to be higher than amongst those who are normal. Likewise, amongst patients with high blood sugar (fbs = 1), the proportion of patients with heart disease would be expected to be higher than what it is amongst normal patients. You can do some data visualization and exploration.

You may want to start with a distribution of status. The following code-chunk will provide you with:

## plotting a histrogram for status
cardio %>%
          ggplot(aes(x = status)) + 
          geom_histogram(stat = 'count', fill = "steelblue") +
          theme_bw()

From this histogram, you can observe that there is almost an equal split between patients having status as normal and abnormal.

This may not always be the case. There might be datasets in which one of the classes in the predicted variable has a very low proportion. Such datasets are said to have a class imbalance problem where one of the classes in the predicted variable is rare within the dataset.

A Credit Card Fraud Detection Model or a Mortgage Loan Default Model are some examples of classification models that are built with a dataset having a class imbalance problem. What other scenarios come to your mind?

You are encouraged to read this article: ROSE: A Package for Binary Imbalanced Learning

You should now explore the distribution of quantitative variables. You can make density plots with frequency counts on the Y-axis and split the plot by the two levels in the status variable.

The following code will produce the plots arranged in a grid of 2 rows

## frequency plots for quantitative variables, split by status  
cardio %>%
  gather(-sex, -cp, -fbs, -restecg, -exang, -slope, -ca, -thal, -status, key = "var", value = "value") %>%
            ggplot(aes(x = value, y = ..count.. , colour = status)) +
            scale_color_manual(values=c("#008000", "#FF0000"))+
            geom_density() +
            facet_wrap(~var, scales = "free",  nrow = 2) +
            theme_bw()

What are your observations from the quantitative plots? Some of your observations might be:

• In all the plots, as we move along the X-axis, the abnormal curve, mostly but not always, lies below the normal curve. You should expect this, as the total number of patients with abnormal is
  smaller. However, for some values on the X-axis (which could be smaller values of X or larger, depending upon the predictor), the abnormal curve lies above.
• For example, look at the age plot. Till x = 55 years, the majority of patients are included in the normal curve. Once x > 55 years, the majority goes to patients
  with
  abnormal and remains so until x = 68 years. Intuitively, age could be a good predictor of status and you may want to partition the data at x = 55 years
  and then again at x = 68 years. When you build your decision tree model, you may expect internal nodes with x > 55 years and x > 68 years.
• Next, observe the plot for chol. Except for a narrow range (x = 275 mg/dl to x = 300 mg/dl), the normal curve always lies above the abnormal curve. You may want to
  form a hypothesis that Cholesterol is not a good predictor of status. In other words, you may not expect chol to be amongst the earliest internal nodes in your decision
  tree model.

Likewise, you can make hypotheses for other quantitative variables as well. Of course, your decision tree model will help you validate your hypothesis.

Now you may want to turn your attention to qualitative variables.

## frequency plots for qualitative variables, split by status  
cardio %>%
       gather(-age, -trestbps, -chol, -thalach, -oldpeak, -status, key = "var", value = "value") %>%
        ggplot(aes(x = value, color = status)) + 
         scale_color_manual(values=c("#008000", "#FF0000"))+
          geom_histogram(stat = 'count', fill = "white") +
          facet_wrap(~var, nrow = 3) +
          facet_wrap(~var, scales = "free",  nrow = 3) +
          theme_bw()

What are your observations from the qualitative plots? How do you want to partition data along the qualitative variables?

• Observe the cp or the chest pain plot. The presence of asymptotic chest pain indicated by cp = 4, could provide a partition in the data and could be among the earliest nodes in your decision tree.
• Likewise, observe the sex plot. Clearly, the proportion of abnormal is much lower (approximately 25%) among females compared to the proportion among males (approximately
  50%). Intuitively, sex might also be a good predictor and you may want to partition the patients’ data along sex. When you build your decision tree model, you may expect internal nodes with sex.

At this point, you may want to go back to both plots and list down the partition (variables and, more importantly, variable values) that you expect to find in your decision tree model.

Of course, all our hypotheses will be validated once we build our decision tree model.

Partitioning data: Training and test sets

Before you start building your decision tree, split the cardio data into a training set and test set:

cardio.train: 70% of the dataset

cardio.test: 30% of the dataset

The following code-chunk will do that:

## Now you can randomly split your data in to 70% training set and 30% test set   
## You should set seed to ensure that you get the same training vs/ test split every time you run the code    
set.seed(1) 

## randomly extract row numbers in cardio dataset which will be included in the training set  
train.index <- sample(1:nrow(cardio), round(0.70*nrow(cardio),0))

## subset cardio data set to include only the rows in train.index to get cardio.train  
cardio.train <- cardio[train.index, ]

## subset cardio data set to include only the rows NOT in train.index to get cardio.test  
## Did you note the negative sign?
cardio.test <- cardio[-train.index,  ]

Classification trees using rpart

“rpart” Package

You will now use rpart package to build your decision tree model. The decision tree that you will build, can be plotted using packages rpart.plot or rattle which provides better-looking plots.

You will use function rpart() to build your decision tree model. The function has the following key arguments:

formula: rpart(, …)

The formula where you declare what predictors you are using in your decision tree. You can specify status ~. to indicate that you want to use all the predictors in your decision tree.

method: rpart(method = < >, …)

The same function can be used to build a decision tree as well as a regression tree. You can use “class” to specify that you are using rpart() function for building a classification tree. If you were building a regression tree, you would specify “anova” instead.

cp rpart(cp = <>,…)

The main role of the Complexity Parameter (cp) is to control the size of the decision tree. Any split that does not reduce the tree’s overall complexity by a factor of cp is not attempted. The default value is  0.01. A value of cp = 1 will result in a tree with no splits. Setting cp to negative values ensures a fully grown tree.

minsplit  rpart( minsplit = <>, …)

The minimum number of observations must exist in a node in order for a split to be attempted. The default value is 20.

minbucket  rpart( minbucket = <>, …)

The minimum number of observations in any terminal node. If only one minbucket or minsplit is specified, the code either sets minsplit to minbucket*3 or minbucket to minsplit/3, which is the default.

You are encouraged to read the package documentation rpart documentation

You can build a decision tree using all the predictors and with a cp = 0.05. The following code chunk will build your decision tree model:

## using all the predictors and setting cp = 0.05 
cardio.train.fit <- rpart(status ~ . , data = cardio.train, method = "class", cp = 0.05)

It is time to plot your decision tree. You can use the function rpart.plot() for plotting your tree. However, the function fancyRpartPlot() in the rattle package is more ‘fancy’

## Using fancyRpartPlot() from "rattle" package
fancyRpartPlot(cardio.train.fit, palettes = c("Greens", "Reds"), sub = "")

Interpreting decision tree plot

What are your observations from your decision tree plot?

Each square box is a node of one or the other type (discussed below):

Root Node cp = 1, 2, 3: The root node represents the entire population or 100% of the sample.

Decision Nodes thal = 3, and ca = 0: These are the two internal nodes that get split up either in further internal nodes or in terminal nodes. There are 3 decision nodes here.

Terminal Nodes (Leaf): The nodes that do not split further are called terminal nodes or leaves. Your decision tree has 4 terminal nodes.

The decision tree plot gives the following information:

Predictors Used in Model: Only the thal, cp, and ca variables are included in this decision tree.

Predicted Probabilities: Predicted probability of a patient being normal or abnormal. Note that the two probabilities add to 100%, at each node.

Node Purities: Each node has two proportions written left and right. The leftmost leaf has 0.82 and 0.18. The number on the left, 0.82 tells you what proportion of the node actually belongs to the predicted class. You can see that this leaf has 82% purity.

Sample Proportion: Each node has a proportion of the sample. The proportion is 100% for the root node. The percentages under the split nodes add up to give the percentage in their parent node.

Predicted class: Each node shows the predicted class as normal or abnormal. It is the most commonly occurring predictor class in that node but the node might still include observations belonging to the other predictor class as well. This forms the concept of node impurity.

Fully grown decision tree

Is this the fully-grown decision tree?

No! Recall that you have grown the decision tree with the default value of cp = 0.05 which ensures that your decision tree doesn’t include any split that does not decrease the overall lack of fit by a factor of 5%.

However, if you change this parameter, you might get a different decision tree. Run the following code-chunk to get the plot of a fully grown decision tree, with a cp = 0

## using all the predictors and setting all other arguments to default 
cardioFull <- rpart(status ~ . , data = cardio.train, method = "class", cp = 0)

## Using fancyRpartPlot() from "rattle" package
fancyRpartPlot(cardioFull, palettes = c("Greens", "Reds"),sub = "")

The fully grown tree adds two more predictors thal and oldpeak to the tree that you built earlier. Now you have seen that changing the cp parameter, gives a decision tree of different sizes – more nodes and/or more leaves. At this stage, you might want to ask the following questions:

• Which of the two decision trees you should go ahead with and present to your division’s Chief Data Scientist? The one developed with a default value of cp = 0.01 or the one with cp = 0?
• Does a bigger decision tree present a better classification model or worse?
• Is the default value of cp = 0.01, the best possible?
• How would you select a cp value that ensures the best-performing decision tree model

There are no thumb rules on how large or small a decision tree should grow. However, you should be aware that:

• A large tree might overfit the data and thus might lead to a model with high variance
• A small tree might miss important parameters and thus might lead to a model with a high bias

So, which of the two decision trees you should present to your division’s Chief Data Scientist? What are the parameters that you can control to build your best decision tree? What are the metrics that you can use to justify the performance of your decision tree model? Conversely, what are the metrics that can help you evaluate the performance of your decision tree model?

Pruning the decision trees

The optimal tree size is chosen adaptively from the training data. The recommended approach is to build a fully-grown decision tree and then extract a nested sub-tree (prune it) in a way that you are left with a tree that has minimal node impurities.

As you have learned in your in-class module, there are three different metrics to calculate the node impurities that can be used for a given node m:

Gini Index:

A measure of total variance across all the classes in the predictor variable. A smaller value of G indicates a purer or more homogeneous node.

Here, Pmk gives the proportion of training observations in the m^th region that are from the k^th class.

Cross-Entropy or Deviance:

Another measure of node impurity:

As with the Gini index, the m^th node is purer if the entropy D is smaller.

In your fitted decision tree model, there are two classes in the predictor variable therefore K = 2 and there are m = 5 regions.

Misclassification Error:

The fraction of the training observations in the m^th node that do not belong to the most common class:

When growing a decision tree, Gini Index or Entropy is typically used to evaluate the quality of the split.

However, for pruning the tree, a Misclassification Error is used.

You can now get back to the fully grown decision tree that you built with cp = 0.

The Complexity Parameter Table will help you evaluate the fitted decision tree model. For your decision tree cardio.train.full, you can print the complexity parameter table using printcp() as well as plot using plotcp()

The CP table will help you select the decision tree that minimizes the misclassification error. CP table lists down all the trees nested within the fitted tree. The best-nested sub-tree can then be extracted by selecting the corresponding value for cp.

The following code will print the CP table for you:

## printing the CP table for the fully-grown tree 
printcp(cardioFull)

Classification tree:
rpart(formula = status ~ ., data = cardio.train, method = "class", 
    cp = 0)

Variables actually used in tree construction:
[1] ca      cp      oldpeak thal    thalach

Root node error: 95/208 = 0.45673

n= 208 

        CP nsplit rel error  xerror     xstd
1 0.536842      0   1.00000 1.00000 0.075622
2 0.063158      1   0.46316 0.52632 0.064872
3 0.031579      3   0.33684 0.38947 0.058056
4 0.015789      4   0.30526 0.35789 0.056138
5 0.000000      6   0.27368 0.36842 0.056794

The plotcp() gives a visual representation of the cross-validation results in an rpart object.

## plotting the cp 
plotcp(cardioFull, lty = 3, col = 2, upper = "splits" )

CP table

How do we interpret the cp table? What is your objective here?

Your objective is to prune the fitted tree i.e. select a nested sub-tree from this fitted tree, such that the cross-validated error or the xerror is the minimum.

The Complexity table for your decision tree lists down all the trees nested within the fitted tree. The complexity table is printed from the smallest tree possible (nsplit = 0 i.e. no splits) to the largest one (nsplit = 8, eight splits). The number of nodes included in the sub-tree is always 1+ the number of splits.

For easier reading, the error columns have been scaled so that the first node (nsplit = 0) has an error of 1. In your decision tree the model with no splits makes 123/267 misclassifications, you can multiply the columns rel error, xerror, and xstd by 123 to get the absolute values. In the first column, the complexity parameter has been similarly scaled. From the cp table we want to select the cp value that minimizes the cross-validated error (xerror).

CP plot

plotcp() gives a visual representation of the CP table. The Y-axis of the plot has the xerrors and the X-axis has the geometric means of the intervals of cp values, for which pruning is optimal. The red horizontal line is drawn 1-SE above the minimum of the curve. A good choice of cp for pruning is typical, the leftmost value for which the mean lies below the red line.

The following code chunk will help you select the best cp from the cp table

## selecting the best cp, corresponding to the minimum value in xerror 
bestcp <- cardioFull$cptable[which.min(cardioFull$cptable[,"xerror"]),"CP"]

## print the best cp
bestcp

0.0157894736842105

You can now use this bestcp to prune the fully-grown decision tree

## Prune the tree using the best cp.
cardio.pruned <- prune(cardioFull, cp = bestcp)

## You can now plot the pruned tree 
fancyRpartPlot(cardio.pruned, palettes = c("Greens", "Reds"), sub = "")   

You can use the summary() function to get a detailed summary of the pruned decision tree. It prints the call, the table shown by printcp, the variable importance (summing to 100), and details for each node (the details depend on the type of tree).

## printing the 
summary(cardio.pruned)  

Call:
rpart(formula = status ~ ., data = cardio.train, method = "class", 
    cp = 0)
  n= 208 

          CP nsplit rel error    xerror       xstd
1 0.53684211      0 1.0000000 1.0000000 0.07562158
2 0.06315789      1 0.4631579 0.5263158 0.06487215
3 0.03157895      3 0.3368421 0.3894737 0.05805554
4 0.01578947      4 0.3052632 0.3578947 0.05613824

Variable importance
      cp     thal    exang  thalach       ca  oldpeak trestbps      age 
      28       17       14       13       12       12        3        2 
     sex 
       1 

Node number 1: 208 observations,    complexity param=0.5368421
  predicted class=normal    expected loss=0.4567308  P(node) =1
    class counts:   113    95
   probabilities: 0.543 0.457 
  left son=2 (109 obs) right son=3 (99 obs)
  Primary splits:
      cp      splits as  LLLR,      improve=34.19697, (0 missing)
      thal    splits as  LRR,       improve=31.59722, (0 missing)
      exang   splits as  LR,        improve=23.76356, (0 missing)
      ca      splits as  LRRR,      improve=21.46291, (0 missing)
      thalach < 147.5 to the right, improve=17.90570, (0 missing)
  Surrogate splits:
      exang   splits as  LR,        agree=0.731, adj=0.434, (0 split)
      thal    splits as  LRR,       agree=0.702, adj=0.374, (0 split)
      thalach < 148.5 to the right, agree=0.683, adj=0.333, (0 split)
      ca      splits as  LRRR,      agree=0.625, adj=0.212, (0 split)
      oldpeak < 0.85  to the left,  agree=0.611, adj=0.182, (0 split)

Node number 2: 109 observations,    complexity param=0.03157895
  predicted class=normal    expected loss=0.1834862  P(node) =0.5240385
    class counts:    89    20
   probabilities: 0.817 0.183 
  left son=4 (98 obs) right son=5 (11 obs)
  Primary splits:
      oldpeak < 1.95  to the left,  improve=5.018621, (0 missing)
      slope   splits as  LRL,       improve=4.913298, (0 missing)
      thal    splits as  LRR,       improve=4.888193, (0 missing)
      ca      splits as  LRRR,      improve=3.642018, (0 missing)
      thalach < 152.5 to the right, improve=3.280350, (0 missing)

Node number 3: 99 observations,    complexity param=0.06315789
  predicted class=abnormal  expected loss=0.2424242  P(node) =0.4759615
    class counts:    24    75
   probabilities: 0.242 0.758 
  left son=6 (35 obs) right son=7 (64 obs)
  Primary splits:
      thal    splits as  LRR,       improve=8.002922, (0 missing)
      exang   splits as  LR,        improve=7.972659, (0 missing)
      ca      splits as  LRRR,      improve=7.539716, (0 missing)
      oldpeak < 0.7   to the left,  improve=3.625175, (0 missing)
      thalach < 175   to the right, improve=3.354320, (0 missing)
  Surrogate splits:
      trestbps < 116   to the left,  agree=0.717, adj=0.200, (0 split)
      oldpeak  < 0.05  to the left,  agree=0.707, adj=0.171, (0 split)
      thalach  < 175   to the right, agree=0.697, adj=0.143, (0 split)
      sex      splits as  LR,        agree=0.677, adj=0.086, (0 split)
      age      < 69.5  to the right, agree=0.667, adj=0.057, (0 split)

Node number 4: 98 observations
  predicted class=normal    expected loss=0.1326531  P(node) =0.4711538
    class counts:    85    13
   probabilities: 0.867 0.133 

Node number 5: 11 observations
  predicted class=abnormal  expected loss=0.3636364  P(node) =0.05288462
    class counts:     4     7
   probabilities: 0.364 0.636 

Node number 6: 35 observations,    complexity param=0.06315789
  predicted class=normal    expected loss=0.4857143  P(node) =0.1682692
    class counts:    18    17
   probabilities: 0.514 0.486 
  left son=12 (20 obs) right son=13 (15 obs)
  Primary splits:
      ca       splits as  LRRR,      improve=7.619048, (0 missing)
      exang    splits as  LR,        improve=6.294925, (0 missing)
      trestbps < 126.5 to the right, improve=2.519048, (0 missing)
      thalach  < 170   to the right, improve=2.057143, (0 missing)
      age      < 53.5  to the left,  improve=1.866667, (0 missing)
  Surrogate splits:
      thalach  < 134   to the right, agree=0.743, adj=0.400, (0 split)
      trestbps < 129   to the right, agree=0.714, adj=0.333, (0 split)
      exang    splits as  LR,        agree=0.686, adj=0.267, (0 split)
      oldpeak  < 1.7   to the left,  agree=0.686, adj=0.267, (0 split)
      age      < 62.5  to the left,  agree=0.657, adj=0.200, (0 split)

Node number 7: 64 observations
  predicted class=abnormal  expected loss=0.09375  P(node) =0.3076923
    class counts:     6    58
   probabilities: 0.094 0.906 

Node number 12: 20 observations
  predicted class=normal    expected loss=0.2  P(node) =0.09615385
    class counts:    16     4
   probabilities: 0.800 0.200 

Node number 13: 15 observations
  predicted class=abnormal  expected loss=0.1333333  P(node) =0.07211538
    class counts:     2    13
   probabilities: 0.133 0.867 

Evaluating decision tree models

You can now use the predict function in rpart package to predict the status of patients included in the test data cardio.test

The following code-chunk predicts the status values for test data and will also print the confusion matrix for actual v/s. predicted values:

## You can now use your pruned tree model to predict the status for your test data 
cardio.predict <- predict(cardio.pruned, cardio.test, type = "class")

You should now evaluate the performance of your model on the test data. You will use your Confusion Matrix and calculate the Classification Error in the predictions:

# confusion matrix (training data)
conf.matrix <- table(cardio.test$status, cardio.predict)
rownames(conf.matrix) <- paste("Actual", rownames(conf.matrix), sep = ":")
colnames(conf.matrix) <- paste("Predicted", colnames(conf.matrix), sep = ":")
print(conf.matrix)

                 cardio.predict
                  Predicted:normal Predicted:abnormal
  Actual:normal                 40                  7
  Actual:abnormal               14                 28

You can calculate the classification error as:

## caclulating the classification error 
round((14 + 7)/89,3)

0.236

So, your decision tree has a 23.6% prediction error. In other words, your model has been able to classify the patients as normal or abnormal with an accuracy of 76.4%. Your division’s Chief Data Scientist should be impressed. Also, you have a classification model that you can very easily explain to doctors.

However, before we wind up, here is a small exercise for you.

Small Exercise:

Decision tree models can suffer from extremely high variance. A small change in the training data can give you very different results. This short exercise is designed to make this point. In the code chunk given below change the values, one at a time, for the following parameters, run the code, and then observe how the decision tree model changes:

set.seed (a): Set the seed to a different number: ‘1234’ or ‘1729’ or ‘9999’ or whatever you like

Training set proportion (p): Set the proportion to different numbers: ‘70%’ or ‘80%’, ‘90%’ or whatever you like

You can go ahead and use the code till the calculation of the prediction error but even plotting the fitted tree would help!

## You should keep the original data frame intact so let's make a copy cardioplay  
cardioplay <- cardio 

## you set the seed to ensure that you get the same training v/s. test split every time you run the code
## Keeping all else constant, you should change the seed from '1234' to any other number 
a <- as.numeric(1234) 


## randomly extract row numbers in cardio dataset which will be included in the training set
## Keeping all else constant, you should change the proportion from '50%' to any other proportion 
p <- as.numeric(0.50)

## You don't need to make any changes in this code-chunk
## Make changes in the code-chunk just above and observe the changes in the output of this code-chunk  

## seed 
set.seed(a) 

## rows in training data 
trainset <- sample(1:nrow(cardioplay), round(p*nrow(cardioplay),0))
cardioplay.train <- cardio[trainset, ]

## rows in test data  
cardioplay.test <- cardio[-trainset,  ] 

## fit the tree 
cardioplay.train.fit <- rpart(status ~ . , data = cardioplay.train, method = "class") 

## plot the tree 
fancyRpartPlot(cardioplay.train.fit, palettes = c("Greens", "Reds"), sub = "")

Conclusion

Now, you have a good understanding of how to perform the exploratory data analysis and prepare your dataset, before you can set out to build a decision tree. You are also familiar with various functions in the rpart package with which you can build decision trees, plot the trees, and prune decision trees to build. As we have discussed earlier, there are other tree-based approaches such as Bagging, Random Forests, and Boosting which improve the accuracy.

You are all set to start practicing exercises on these advanced topics!

 #Bike sharing data set as input to the module 
dataset <-maill.mapInputPort(1) 
#extracting hour, weekday, month, and year fromthe  dataset
dataset$datetime <- as.POSIXct(dataset$datetime, format = "%m/%d/%Y %I:%M:%S %p")
dataset$hour  <- substr(dataset$datetime,	12,13)
dataset$weekday  <- weekdays(dataset$datetime)
dataset$month  <- months(dataset$datetime)
dataset$year  <- substr(dataset$datetime,	1,4)
#Preserving the column order 
Count <- dataset[,names(dataset) %in% c("count")]
 OtherColumns <- dataset[,!names(dataset) %in% c("count")]
dataset <- cbind(OtherColumns,Count)
#Remova e single observation with weather = 4 to preventhe t scoring model from failing
dataset <- subset(dataset, weather != '4')

#Return the dataset after appending the new featuresmailml.mapOutputPort("dataset");

#Trainer R Script
#Input: dataset
#Output: model
library(randomForest)
model <- randomForest(Count ~ ., dataset)
 #Scorer R Script
#Input: model, dataset
#Output: scores
library(randomForest)
scores <- data.frame(predict(model, subset(dataset, select = -c(Count))))
names(scores) <- c("Predicted Count")

#Results as input to module
dataset1 <- maml.mapInputPort(1)
countMSE <- mean((dataset1$Count-dataset1["Predicted Count"])^2)
evaluation <- data.frame(countMSE)
#Output evaluation
maml.mapOutputPort("evaluation");

Written by Phuc Duong

Data Science Dojo has launched Jupyter Hub for Machine Learning using Python offering to the Azure Marketplace with pre-installed machine learning libraries and pre-cloned GitHub repositories of famous machine learning books which help the learner to take the first steps into the field of machine learning.

What is machine learning?

Machine learning is a sub-field of Artificial Intelligence. It is an innovative technology that allows machines to learn from historical data and provide the best results to predict outcomes.

Machine learning using Python

Machine learning requires exploratory data analysis, data processing, and the training of data to predict outcomes. Python provides a vast number of libraries and frameworks that let the user collect, analyze and transform data by just using built-in functions provided by the library which makes coding easy and also saves a significant amount of time.

 PRO TIP: Join our 5-day instructor-led Python for Data Science training to enhance your machine learning skills.

Challenges for individuals

Individuals who are new to machine learning and want to excel in their path in machine learning usually lack computing as well as learning resources to gain hands-on experience with machine learning. A beginner in machine learning also faces compatibility issues while installing libraries.

What we provide

With just a single click, Jupyter Hub for Machine Learning using Python comes with pre-installed machine learning python libraries, which gives the learner an effortless coding environment in the Azure cloud and reduces the burden of installation. Moreover, this offer provides the learner with repositories of famous books on machine learning which contain chapter-wise notebooks which serve as a learning resource for a user in gaining hands-on experience with machine learning. The heavy computations required for Machine Learning applications are not performed on the user’s local machine. Instead, they are performed in the Azure cloud, which increases responsiveness and processing speed.

Listed below are the pre-installed machine learning python libraries and the sources of repositories of machine learning books provided by this offer:

Python libraries

• Pandas
• NumPy
• scikit-learn
• mlpack
• matplotlib
• SciPy
• Theano
• Pycaret
• Orange3
• seaborn

Repositories

•  Github repository of book ‘Python Machine Learning Book 1st Edition’, by author Sebastian Raschka.
•  Github repository of book ‘Python Machine Learning Book 2nd Edition’, by author Sebastian Raschka.
•  Github repository of the book ‘Hands-on Machine Learning with Scikit Learn, Keras, and TensorFlow’, by author Geron-Aurelien.
•  Github repository of ‘Microsoft Azure Cloud Advocates 12-week Machine Learning curriculum’.

Conclusion

Jupyter Hub for Machine Learning using Python provides an in-browser coding environment with just a single click, hence providing ease of installation. Through this offer, a user can work on a variety of machine learning applications including stock market trading, email spam and malware filtering, product recommendations, online customer support, medical diagnosis, online fraud detection, and image recognition.

Jupyter Hub for Machine Learning using Python offered by Data Science Dojo is ideal to learn more about machine learning without the need to worry about configurations and computing resources. The heavy resource requirement for processing and training large data for these applications is no longer an issue as data-intensive computations are now performed on Microsoft Azure which increases processing speed.

At Data Science Dojo, we deliver data science education, consulting, and technical services to increase the power of data. We are therefore adding a free Jupyter Notebook Environment dedicated specifically for Machine Learning using Python. The offering leverages the power of Microsoft Azure services to run effortlessly with outstanding responsiveness. Install the Jupyter Hub offer now from the Azure Marketplace by Data Science Dojo, your ideal companion in your journey to learn data science!

Learn the difference between supervised ML, unsupervised ML, and reinforcement learning. Test your knowledge of machine learning techniques with an interactive infographic.

Have you noticed that we have two machine learning demos on our site that allow you to deploy predictive models?

The Titanic Survival Predictor is designed to work with a Microsoft Azure model for machine learning. The AWS Machine Learning Caller is our new demo that connects to an Amazon Machine Learning model.

You can use Microsoft Azure ML or Amazon ML to build your machine learning model, but what’s the difference between the two approaches?

The idea is that you can use Microsoft Azure ML or Amazon ML to build a machine-learning model, and then use our demo to input values for the prediction.

Each ML program provides an endpoint that you can use to access the model and run predictions. Our demos interface with that endpoint and provide a graphic user interface for making predictions.

So, what’s the difference between the machine learning demos?

First of all, the backend is different. But we’ll keep this brief.

The graphic below shows what types of models can be run through the demo.

• The cruise ship represents the Titanic classification model generated from our Azure ML tutorial.
• The iris represents any classification model, such as a model used to predict species from a set of measurements.
• The complicated graph represents a regression model. Regression models are used to predict a number given a set of input numbers.

You can see that the Titanic model can link to both demos, but the classification (iris) model only links to our Amazon demo. The numerical dataset does not work with either of our demos.

The demos are currently limited to classification models only (because linear regression models work differently and require a different backend).

MLaaS: User perspectives

From the user’s perspective, the Titanic Survival Predictor is built for a specific purpose. It interfaces with the exact Titanic classification model that we created for Azure and is included as part of our bootcamp. Users can change all the tuning parameters and make the model unique.

However, the input variables, or “schema” to be labeled the same way as the original model or it won’t work.

So, if you rename one of the columns, the demo will have an error. However, since we published the Azure model online, it’s pretty easy to copy the model and change some parameters.

To get your predictive model to work with our Titanic Survival Predictor demo, you’ll need the following information:

• Name (used to generate your own url)
• Post URL (or endpoint)
• API key

The AWS Machine Learning Caller is not built for a specific dataset like Titanic. It will work with any logistic regression model built in Amazon Machine Learning. When you input your access keys and model id, our demo automatically pulls the schema from Amazon.

It does not require a specific schema like our Titanic Survival Predictor.

To get your predictive model to work with our AWS Machine Learning Caller demo, you’ll need the following information:

• Access key
• Secret access key
• AWS Account Region
• AWS ML Model ID

Why do two machine learning demos do similar things?

These are training tools for our 5-day bootcamp. We use Microsoft Azure to teach classification models. The software has tools for data cleaning and manipulation. The way that the tools are laid out is visual and easy to understand. It provides a clear organization of the processes: input data, clean data, build a model, evaluate the model, and deploy the model.

Microsoft Azure has been a great way to teach the model-building process.

We’ve recently added Amazon Machine Learning to our curriculum. The program is simpler, where all the processes described above are automated. Amazon ML walks users through the process.

However, it does provide slightly different evaluation metrics than Microsoft Azure, so we use it to teach regression and classification models as well.

Help us get better!

We are always looking for ways to incorporate new tools into our curriculum. If there is a tool that you think we ought to have, please let us know in the comments.

Or, you can contact us here

Written by Phuc Duong

This Azure tutorial will walk you through deploying a predictive model in Azure Machine Learning, using the Titanic dataset.

The classification model, covered in this article, uses the Titanic dataset to predict whether a passenger will live or die, based on demographic information. We’ve already built the model for you and the front-end UI. This tutorial will show you how to customize the Titanic model we built and deploy your own version.

MLaaS overview:

About the data

The Titanic dataset’s complexity scales up with feature engineering, making it one of the few datasets good for both beginners and experts. There are numerous public resources to obtain the Titanic dataset, however, the most complete (and clean) version of the data can be obtained from Kaggle, specifically their “train” data.

The “train” Titanic data ships with 891 rows, each one about a passenger on the RMS Titanic, the night of the disaster. The dataset also has 12 columns that record attributes of each passenger’s circumstances and demographics such as passenger id, passenger class, age, gender, name, number of siblings and spouses aboard, number of parents and children aboard, fare, ticket number, cabin number, port of embarkation, and whether or not they survived.

For additional reading, a repository of biographies about everyone aboard the RMS Titanic can be found here (complete with pictures).

Getting the experiment

About the Titanic Survival User Interface

From the dataset, we will build a predictive model and deploy the model in AzureML as a web service. Data Science Dojo has built a front-end UI to interact with such a web service.

Click on the link below to view a finished version of this deployed web service.

Titanic Survival Predictor

Use the app to see what your chance of survival might have been if you were on the Titanic. Play around with the different variables. What factors does the model deem important in calculating your predicted survival rate?

The following tutorial will walk you through how to deploy a titanic prediction model as a web service.

Get an Azure ML account

This MLaaS tutorial assumes that you already have an AzureML workspace. If you do not, please visit the following link for a tutorial on how to create one.

Creating Azure ML Workspace

Please note that an Azure ML 88-hourfree trial does not have the option of deploying a web service.

Clone the experiment

For this MLaaS tutorial, we will provide you with the completed experiment by letting you clone ours. If you are curious about how we created the experiment, please view our companion tutorial where we talk about where we talk about the process of data mining.

Our experiment is hosted in the Azure ML public gallery. Navigate to the experiment by clicking on the link below or by clicking “Clone ont to Azure ML” within the Titanic Survival Predictor web page itself. The Azure ML Gallery is a place where people can showcase their experiments within the Azure ML community.

Gallery Titanic Experiment

Click on the “open in studio” button.

The experiment and dataset will be copied to your studio workspace. You should now see a bunch of modules linked together in a workflow. However, since we have not run the experiment, the workflow is only a set of instructions that Azure ML will use to build your models. We will have to run the experiment to produce anything.

Click the “run” button at the bottom middle of the AzureML window.

This will execute the workflow that is present within the experiment. The experiment will take about 2 minutes and 30 seconds to finish running. Wait until every module has a green checkmark next to it. This indicates that each module has finished running.

MLaaS predictive model evaluation and deployment

Select an algorithm

You may have noticed that the cloned experiment shipped with two predictive models–two different decision forests. However, because we can only deploy one predictive model, we should see which performs better. Right click on the output node of the evaluate model module and click “visualize.”

Evaluate your model

For the purpose of this tutorial, we will define the “better” performing model as the one that scored a higher RoC AuC. We will gloss over evaluating performance metrics of classification models since that would require a longer, more in-depth discussion.

In the evaluate model module, you will see an “ROC” graph with a blue and red line graphed on it. The blue line represents the RoC performance of the model on the left and the red line represents the performance of the model on the right.

The higher the curve is on the graph, the better the performance. Since the red curve, the right model is higher on the graph than the blue curve, we can say that the right model is the better-performing model in this case. We will now deploy the corresponding decision tree model.

Deploy the experiment

Before deployment, all modules must have a green check mark next to them.

To deploy the selected decision forest model, select the “train model module” on the right.

While that is selected, hover over the “setup web service” button on the bottom middle of the screen. A pull-up menu will appear. Select “predictive web service”.

Azure ML will now remove and consolidate unnecessary modules, then it will automatically save the predictive model as a trained model and set up web service inputs and outputs.

Drop the response class

Our web service is almost complete. However, we need to tune the logic behind the web service function. The score model module is the module that will execute the algorithm against a given dataset. The score model module can also be called the “prediction module” because that is what happens when you apply a trained algorithm against a dataset.

You will notice that the score model module also takes in a dataset on the right input node. When deploying a predictive model, the score model module will need a copy of the required schema. The dataset used to train the model is fed back into the score model module because that is the schema that our trained algorithm currently knows.

However, that schema also holds our response class “survived,” the attribute that we are trying to predict. We must now drop the survived column. To do this we will use the “project columns” module. Search for it in the search bar on the left side of the AzureML window, then drag it into the workspace.

Replicate the picture on the left by connecting the last metadata editor’s output node to the input of the new project columns module. Then connect the output of the new project columns module with the right input of the score model module.

Select the project columns module once the connections have been made. A “Properties” window will appear on the right side of the AzureML window. Click on “launch column selector.”

To drop the “Survived” column we will “Begin with: All Columns,” then choose to “Exclude” by “column names,” “Survived.”

Reroute web service input

We must now point our web service input in the correct direction. The web service input is currently pointing to the beginning of the workflow where data was cleaned, columns were renamed, and columns were dropped. However, the form on the Titanic Prediction App will do the cleansing for you.

Let’s reroute the web service input to point directly at our score model module. Drag the web service input module down toward the score model module and connect it to the right input node of the score model (the same node that the newly added project columns module is also connected to).

Deploy your model

Once all the rerouting has been done, run your experiment one last time. A “Deploy Web Service” button should now be clickable at the bottom middle of the Azure ML window. Click this and AzureML will automatically create and host your web service API with your own endpoints and post-URL.

Exposing the deployed webservice

Test your web service

You should now be on the web deployment screen for your web service. Congratulations! You are now in possession of a web service that is connected to a live predictive model. Let’s test this model to see if it behaves properly.

Click the “test” button in the middle of the web deployment screen. A window with a form should pop up. This form should look familiar because it is the same form that the Titanic Predictor App was showing you.

Send the form a few values to see what it returns. The predictions will come in JSON format. The last number in JSON is the prediction itself, which should be a decimal akin to a percentage. This percentage is the predicted likelihood of survival based upon the given parameters, or in this case the passenger’s circumstances while aboard the Titanic.

Find your API key

The API key is located on the web deployment screen, above the test button that you clicked on earlier. The API key input box comes with a copy to clipboard button, click on that button to copy the key. Paste the key into the “Add Your Own Model” page.

Get your post URL

To grab the post-URL, click on the “REQUEST/RESPONSE” button, to the left of the test button. This will take you to the API help page.

Under “Request” and to the right of “POST” is the URL. Copy and paste this URL into the “Add Your Own Model” form.

Enjoy and share

You now have your very own web service! Remember to save the URL because it is your own web page that you may share with others.

If you have a free trial Azure ML account please note that your web service may discontinue when your free trial subscription ends.

Written by Phuc Duong

All of these written texts are unstructured; text mining algorithms and techniques work best on structured data.

Text analytics for machine learning: Part 1

Have you ever wondered how Siri can understand English? How can you type a question into Google and get what you want?

Over the next week, we will release a five-part blog series on text analytics that will give you a glimpse into the complexities and importance of text mining and natural language processing.

This first section discusses how text is converted to numerical data.

In the past, we have talked about how to build machine learning models on structured data sets. However, life does not always give us data that is clean and structured. Much of the information generated by humans has little or no formal structure: emails, tweets, blogs, reviews, status updates, surveys, legal documents, and so much more. There is a wealth of knowledge stored in these kinds of documents which data scientists and analysts want access to. “Text analytics” is the process by which you extract useful information from text.

Some examples include:

• Predicting the stock market with Tweets
• Detecting fraudulent activities from email messages
• Forecast box office success from just the screen play
• Evaluating cultural fit and personality type from resumes

All these written texts are unstructured; machine learning algorithms and techniques work best (or often, work only) on structured data. So, for our machine learning models to operate on these documents, we must convert the unstructured text into a structured matrix. Usually this is done by transforming each document into a sparse matrix (a big but mostly empty table). Each word gets its own column in the dataset, which tracks whether a word appears (binary) in the text OR how often the word appears (term-frequency). For example, consider the two statements below. They have been transformed into a simple term frequency matrix. Each word gets a distinct column, and the frequency of occurrence is tracked. If this were a binary matrix, there would only be ones and zeros instead of a count of the terms.

Make words usable for machine learning

Why do we want numbers instead of text? Most machine learning algorithms and data analysis techniques assume numerical data (or data that can be ranked or categorized). Similarity between documents is calculated by determining the distance between the frequency of words. For example, if the word “team” appears 4 times in one document and 5 times in a second document, they will be calculated as more similar than a third document where the word “team” only appears once.

Text mining: Build a matrix

While our example was simple (6 words), term frequency matrices on larger datasets can be tricky.

Imagine turning every word in the Oxford English dictionary into a matrix, that’s 171,476 columns. Now imagine adding everyone’s names, every corporation or product or street name that ever existed. Now feed it slang. Feed it every rap song. Feed it fantasy novels like Lord of the Rings or Harry Potter so that our model will know what to do when it encounters “The Shire” or “Hogwarts.” Good, now that’s just English. Do the same thing again for Russian, Mandarin, and every other language.

After this is accomplished, we are approaching a several billion-column matrix; two problems arise. First, it becomes computationally unfeasible and memory intensive to perform calculations over this matrix. Secondly, the curse of dimensionality kicks in and distance measurements become so absurdly large in scale that they all seem the same. Most of the research and time that goes into natural language processing is less about the syntax of language (which is important) but more about how to reduce the size of this matrix.

Now that we know what we must do and the challenges that we must face in order to reach our desired result. The next three blogs in the series will be directly addressed to these problems. We will introduce you to 3 concepts: conforming, stemming, and stop word removal.

Want to learn more about text mining and text analytics?

Check out our short video on our data science bootcamp curriculum page OR watch our video on tweet sentiment analysis.

Written by Phuc Duong

Develop an understanding of text analytics, text conforming, and special character cleaning. Learn how to make text machine-readable.

Text analytics for machine learning: Part 2

Last week, in part 1 of our text analytics series, we talked about text processing for machine learning. We wrote about how we must transform text into a numeric table, called a term frequency matrix, so that our machine learning algorithms can apply mathematical computations to the text. However, we found that our textual data requires some data cleaning.

In this blog, we will cover the text conforming and special character cleaning parts of text analytics.

Understand how computers read text

The computer sees text differently from humans. Computers cannot see anything other than numbers. Every character (letter) that we see on a computer is actually a numeric representation to a computer, with the mapping between numbers and characters determined by an “encoding table.” The simplest, but most common, is ASCII encoding in text analytics. A small sample ASCII table is shown to the right.

Any human would know that this is just six different spellings for the same word, but to a computer these are six different words. These would spawn six different columns in our term-frequency matrix. This will bloat our already enormous term-frequency matrix, as well as complicate or even prevent useful analysis.

Unify words with the same spelling

To unify the six different “CAFÉ’s”, we can perform two simple global transformations.

Casing: First we must convert all characters to the same casing, uppercase or lowercase. This is a common enough operation. Most programming languages have a built-in function that converts all characters into a string into either lowercase or uppercase. We can choose either global lowercasing or global uppercasing, it does not matter as long as it’s applied globally.

String normalization: Second, we must convert all accented characters to their unaccented variants. This is often called Unicode normalization, since accented and other special characters are usually encoded using the Unicode standard rather than the ASCII standard. Not all programming languages have this feature out of the box, but most have at least one package which will perform this function.

Note that implementations vary, so you should not mix and match Unicode normalization packages. What kind of normalization you do is highly language dependent, as characters which are interchangeable in English may not be in other languages (such as Italian, French, or Vietnamese).

Remove special characters and numbers

The next thing we have to do is remove special characters and numbers. Numbers rarely contain useful meaning. Examples of such irrelevant numbers include footnote numbering and page numbering. Special characters, as discussed in the string normalization section, have a habit of bloating our term-frequency matrix. For instance, representing a quotation mark has been a pain-point since the beginning of computer science.

Unlike a letter, which may only be capital or not capital, quotation marks have many popular representations. A quotation character has three main properties: curly, straight, or angled; left or right; single, double, or triple. Depending on the text analytics encoding used, not all of these may exist.

The table below shows how quoting the word “café” in both straight quote and left-right quotes would look in a UTF-8 table in Arial font.

Avoid over-cleaning

The problem is further complicated by each individual font, operating system, and programming language since implementation of the various encoding standards is not always consistent. A common solution is to simply remove all special characters and numeric digits from the text. However, removing all special characters and numbers can have negative consequences.

There is a thing as too much data cleaning when it comes to text analytics. The more we clean and remove the more “lost in translation” the textual message may become. We may inadvertently strip information or meaning from our messages so that by the time our machine learning algorithm sees the textual data, much or all the relevant information has been stripped away.

For each type of cleaning above, there are situations in which you will want to either skip it altogether or selectively apply it. As in all data science situations, experimentation and good domain knowledge are required to achieve the best results.

When do we want to avoid over-cleaning in your text analytics?

Special characters: The advent of email, social media, and text messaging have given rise to text-based emoticons represented by ASCII special characters.

For example, if you were building a sentiment predictor for text, text-based emoticons like “=)” or “>:(” are very indicative of sentiment because they directly reference happy or sad. Stripping our messages of these emoticons by removing special characters will also strip meaning from our message.

Numbers: Consider the infinitely gridlocked freeway in Washington state, “I-405.” In a sentiment predictor model, anytime someone talks about “I-405,” more likely than not the document should be classified as “negative.” However, by removing numbers and special characters, the word now becomes “I”. Our models will be unable to use this information, which, based on domain knowledge, we would expect to be a strong predictor.

Casing: Even cases can carry useful information sometimes. For instance, the word “trump” may carry a different sentiment than “Trump” with a capital T, representing someone’s last name.

One solution to filter out proper nouns that may contain information is through name entity recognition, where we use a combination of predefined dictionaries and scanning of the surrounding syntax (sometimes called “lexical analysis”). Using this, we can identify people, organizations, and locations.

Next, we’ll talk about stemming and Lemmatization as a way to help computers understand that different versions of words can have the same meaning (ex. run, running, runs).

Learn more

Want to learn more about text analytics? Check out the short video on our curriculum page OR

Learn how companies like Zillow predict the value of your home. Build a predictive model using Azure machine learning that estimates the real estate sales price of a house.

Ames housing dataset includes 81 features and 1460 observations. Each observation represents the sale of a home and each feature is an attribute describing the house or the circumstance of the sale.

Clone this experiment to build a predictive model

A full copy of this experiment has been posted to the Cortana Intelligence Gallery. Go to the link and click on “open in Studio.”

Preprocessing & data exploration

Drop low-value columns

Begin by identifying features (columns) that add little to no value for predictive modeling. These columns will be dropped using the “select columns from dataset” module.

The following columns were chosen to be “excluded” from the dataset:

Id, Street, Alley, PoolQC, Utilities, Condition2, RoofMatl, MiscVal, PoolArea, 3SsnPorch, LowQualFinSF, MiscFeature, LandSlope, Functional, BsmtHalfBath, ScreenPorch, BsmtFinSF2, EnclosedPorch.

These low-quality features were removed to improve the model’s performance. Low quality includes lack of representative categories, too many missing values, or noisy features.

Define categorical variables

We must now define which values are non-continuous by casting them as categorical. Mathematical approaches for continuous and non-continuous values differ greatly. Nominal categorical features were identified and cast to categorical data types using the metadata editor to ensure proper mathematical treatment by the machine learning algorithm.

The first edit metadata module will cast all strings. The column “MSSubClass” uses numeric integer codes to represent the type of building the house is, and therefore should not be treated as a continuous numeric value but rather a categorical feature. We will use another metadata editor to cast it into a category.

Clean missing data

Most algorithms are unable to account for missing values and some treat it inconsistently from others. To address this, we must make sure our dataset contains no missing, “null,” or “NA” values.

Replacement of missing values is the most versatile and preferred method because it allows us to keep our data. It also minimizes collateral damage to other columns due to one cell’s bad behavior. Numerical values can easily be replaced with statistical values such as mean, median, or mode.

While categories can be commonly dealt with by replacing with the mode or a separate categorical value for unknowns.

For simplicity, all categorical missing values were cleaned with the mode and all numeric features were cleaned using the median. To further improve a model’s performance, custom cleaning functions should be tried and implemented on each individual feature rather than a blanket transformation of all columns.

Machine learning – Model building

Statistical feature selection

Not every feature in its current form is expected to contain predictive value to the model and may mislead or add noise to the model. To filter these out we will perform a Pearson correlation to test all features against the response class (sales price) as a quick measure of their predictive strength, only picking the top X strongest features from this method, the remaining features will be left behind.

This number can be tuned for further model performance increases.

Select an algorithm

First, we must identify what kind of machine learning problem this is: classification, regression, clustering, etc. Since the response class (sales price) is a continuous numeric value, we can tell that it is a regression problem. We will use a linear regression model with regularization to reduce the over-fitting of the model.

• To ensure a stable convergence of weight and biases, all features except the response class must be normalized to be placed into the same range.

Model training and evaluation

The method of cross-validation will be used to evaluate the predictive performance of the model as well as that performance’s stability in regard to new data. Cross-validation will build ten different models on the same algorithm but with different and non-repeating subsets of the same dataset. The evaluation metrics on each of the ten models will be averaged and a standard deviation will infer the stability of the average performance.

Parameter tuning

This experiment will build a regression model that minimizes the mean RMSE of the cross-validation results with the lowest variance possible(but also considers bias-variance trade-offs).

1. The first regression model was built using default parameters and produced a very inaccurate model ($124,942 mean RMSE) and was very unstable (11,699 standard deviation).
2. The high bias and high variance of the previous model suggest the model is over-fitting to the outliers and is under-fitting the general population.
3. The L2 regularization weight will be decreased to lower the penalty of higher coefficients. After lowering the L2 regularization weight, the model is more accurate with an average cross-validation RMSE of $42,366.
4. The previous model is still quite unstable with a standard deviation of $8,121. Since this is a dataset with a small number of observations (1460), it may be better to increase the number of training epochs so that the algorithm has more passes to reach convergence.
5. This will increase training times but also increase stability. The third linear model had the number of training epochs increased and saw a better mean cross-validation RMSE of $36,684 and a much more stable standard deviation of $3,849.
6. The final model had a slight increase in the learning rate which improved both mean cross-validation RMSE and the standard deviation.

Deployment

The algorithm parameters that yielded the best results will be the ones that are shipped. The best algorithm (the last one) will be retrained using 100% of the data since cross-validation leaves 10% out each time for validation.

Further improvement of this Azure machine learning model

Feature engineering was entirely left out of this experiment. Try engineering more features from the existing dataset to see if the model will improve. Some columns that were originally dropped may become useful when combined with other features. For example, try bucketing the years in which the house was built by decade. Clustering the data may also yield some hidden insights.

Written by Phuc Duong

If you’re an aspiring data scientist, you will need to know some math. Learn why math is important in machine learning.

At the end of each of our bootcamps we ask our students to provide us with feedback on their experience. In particular, we ask for honest assessments and opinions on how we can improve. It’s something we take very seriously at Data Science Dojo and I can list a number of changes we’ve made as a direct result of student feedback. Given that our students come from a broad spectrum of backgrounds, it is not surprising that we invariably receive feedback that distills down to, “why so much mathematics for machine learning?”

Mathematics, machine learning, and programming- The tools of the data scientist

It is my firm conviction that you do not need a PhD in Statistics/Computer Science/Machine Learning/Whatever to become a Data Scientist. However, it is my firm conviction that ultimately Data Science boils down to two things – Mathematics and Programming. Per this belief, our Bootcamp curriculum is engineered to provide the required foundation in both mathematical concepts/theory and programming for Data Science.

As you might imagine, it is exceedingly rare for our students to provide feedback along the lines of, “Why so much programming?” Some students comment that there was more programming than they expected, but rarely is the need for a Data Scientist to have coding skills questioned. Not so for mathematics. This is unfortunate as I would strenuously argue that without some mathematical knowledge, a Data Scientist will not be able to build effective models.

Here’s a list of Techniques for Data Scientists to Upskill with LLMs

A hypothetical example

Here’s a hypothetical example to illustrate my point. An aspiring Data Scientist does some research regarding a particular problem and finds a blog post, a paper, and/or a forum post recommending the application of a regression model built with Stochastic Gradient Descent to the problem space. The following screenshot is an excerpt from Python’s most excellent scikit-learn library.

NOTE – Rest assured that similar R examples exist as well (e.g., the awesome glmnet pacakge) and I only use scikit-learn here as the scikit-learn HTML documentation is more visually attractive ;-).

The above green boxes illustrate some of the mathematical knowledge required to use this algorithm to build the most effective model. For example:

1. The Stochastic Gradient Descent algorithm – what is it and how does it work.
2. Regularization – what is it and how does it work.
3. The differences between L1 and L2 regularization – why a Data Scientist might want one vs. the other or a blend of both.

This simple example illustrates my point about math and programming. Specifically, this example shows that without the required math knowledge, the Data Scientist has little hope of coding up the training/construction of the most effective model in any reasonable way.

The mathematics takeaway

For these reasons, our students learn every highlighted item above as part of our curriculum’s coverage of regression. We also teach our students mathematics and theory for other important topics like decision trees, boosting, and recommender systems. It is also for these reasons that I advise the aspiring Data Scientists that I mentor that eventually, they will need to dust off their math textbooks.

Until next time, happy data sleuthing!

Written by Dave Langer

It’s not easy going through thousands of SEC filings when you’re part of that SASB. Sustainability data and machine learning can make that job easier.

Imagine being an investor before Volkswagen’s recent emissions scandal. Now imagine pricing in the risk of Volkswagen’s governance controls relative to their peers before that. Or, imagine being able to account for Chipotle’s food safety risks in their supply chain before their issues with E.coli in recent years.

Hindsight is 20/20, yes. But there were signs in both cases from their public Securities and Exchange Commission (SEC) filings that these risks were evident relative to their peers. There were also signs that sustainability data or Environmental Social Governance (ESG) data could have made this more transparent to others.

Too often, ‘sustainability’ is often associated with environmental issues. Sustainability data also encompasses issues related to company self-governance and company product safety. In fact, the first ESG quantitative investment fund, Arabesque Partners, has been using this type of ESG data to exclude companies (read more about Arabesque Partners here). In their recent case study, they mention that they use this data to  “not include Toshiba, Valiant and Volkswagen.”[1] These are just a few examples of companies they were able to identify with ESG risks.

Yes, the benefits of sustainability data have become more mainstream. Although, there is still a lot of this data locked away in unstructured text form in a company’s SEC filings. One can’t simply read through an industry or sector’s worth of lengthy disclosures, along with that of their peers to find these differences and compare them on an apples-to-apples basis.

This becomes more difficult when you consider trying to read the text for all of these companies and then classify them into decision-useful data for sustainability, or sustainability data. Important questions arise:

• What topic or category would you come up with?
• How would you know those categories are important?
• How would you group companies? By industry? By sector?
• What classification system would you use and would that apply to sustainability issues?

This is where the work of my organization, the Sustainability Accounting Standards Board (SASB), has provided guidance for what topics may likely be material (relevant) for a company. This is based on its industry classification within our SICS company classification system.  Our mission has been to provide industry-specific sustainability standards based on exhaustive research and industry working group participation. They focus disclosure on what is likely to be material and relevant.

Our internal SASB research team did exactly this work of classifying corporate SEC disclosures on just a small sample of 5-10 companies and industries. It has been very useful to get this data. Unfortunately, it has taken over 2 years plus a team of researchers to get even a fraction of the total economy of companies. While sampling can attempt to represent the overall distribution of the sustainability data, we knew from our experiences interacting with external stakeholders that a large amount of untapped, valuable data existed.

The goal

We realized that we needed a way to look at the tens of thousands of companies that make annual filings with the SEC. This is in order to find a way to measure qualitative text disclosure. This way we could show the changes in disclosure over time on SASB’s sustainability standards. In the words of senior management, creating a way to measure corporate sustainability data disclosure at scale in SEC filings would be “SASB’s performance metric for success” and show our impact on society.

We reasoned that by showing existing disclosure on our topics, this could better incentivize companies to improve not only disclosure but actual management of these issues because of heightened attention from investors, regulators, and other key stakeholders. In the spirit of Justice Brandeis’ philosophy on transparency, disclosure would bring the “sunlight” to parts of corporate disclosure that investors and the general public would not be able to find without our standards and lens for finding this sustainability data.

The solution: Sustainability data

We chose to look into machine learning as a way to scale our efforts. We started our project as a small pilot with just one sector. This was to see the feasibility of this project and we had the assistance of two amazing data science contractors.

However, we found that bigger challenges awaited us with scaling this effort from a single sector to the full economy of 10 sectors that SASB has standards for. As the program manager for our efforts with this project, I realized that I needed deeper and practical understanding of data science. This was in order to make better decisions related to the design and implementation of this pipeline.

Now I had taken some data science courses online and gained practical experience working with our data science contractors before. But to run the pilot program, I realized that I needed more training and exposure on how to apply machine learning in practice.

Just five short months ago, I was fortunate enough to be accepted as the first nonprofit fellow for Data Science Dojo. In one short week I was able to build a data pipeline end-to-end, compete in my first Kaggle contest, the crowd-sourced data science platform, (read more here), and develop a core fundamental understanding of the core steps of building a classifier.

Before this bootcamp, I had taken plenty of data science courses and had worked with data scientists but it was hard to connect the pieces from data exploration & analysis and feature engineering to developing performance metrics and testing different hyper-parameters.

After the bootcamp, I went back to my organization understanding how to modify existing parts of our machine learning pipeline to scale and also integrate it with manual components such as Mechanical Turk. Using skills I learned at the Data Science bootcamp, I was able to tweak parameters of our classifier and test an entirely different approach to how to approach semi-supervised learning. In particular, I was able to use my feature engineering skills to create and select features that I would not have considered for the classifier. In addition, I could augment our process by building checkpoints that were able to help with data quality validation. I attribute this to the fact that I could now see how things were all connected.

Without the fundamentals that I learned at the Data Science Dojo bootcamp, it would have been challenging to be at the point we are at today. We are nearing the completion of a machine learning pipeline with almost 1M classified excerpts and a corresponding web application displaying this data that we will launch to the public in the fall.

Our hope is that this data will change the way capital markets view sustainability and that investors will be able to use this sustainability data to influence the decisions that are made by companies in regard to the material sustainability issues that my organization has researched. This can help to shift the allocation of funds from organizations that focus on less sustainable outcomes to organizations that account for the greatest challenges in climate change, air quality, water management, hazardous materials, material sourcing, and other important sustainability issues.

Contributors:

Michael D’Andrea: As the former data science program manager for SASB, Michael managed and analyzed large, diverse and unstructured sustainability datasets for trends that support the greater disclosure of sustainability information for the public. He has a M.A. in computer education and was a data science dojo fellow.

Sabrina Dominguez: Sabrina holds a B.S. in Business Administration with a specialization in Marketing Management from Central Washington University. She has a passion for search engine optimization and marketing.

This tutorial will walk you through building a classification model in Azure ML Studio by using the same process as a traditional data mining framework.

Using Azure ML studio (Overview)

We will use the public Titanic dataset for this tutorial. From the dataset, we can build a predictive model that will correctly classify whether you will live or die based upon a passenger’s demographic features and circumstances.

Would you survive the Titanic disaster?

About the data

We use the Titanic dataset in our data science bootcamp, and have found it is one of the few datasets that is good for both beginners and experts because its complexity scales up with feature engineering. There are numerous public resources to obtain the Titanic dataset, however, the most complete (and clean) version of the data can be obtained from Kaggle, specifically their “train” data.

The train Titanic data has 891 rows, each one pertaining to a passenger on the RMS Titanic on the night of its disaster. The dataset also has 12 columns that each record an attribute about each occupant’s circumstances and demographics: user ID, passenger class, age, gender, name, number of siblings and spouses aboard, number of parents and children aboard, fare price, ticket number, cabin number, their port of embarkation, and whether they survived the ordeal or not.

For additional reading, a repository of biographies pertaining to everyone aboard the RMS Titanic can be found here (complete with pictures).

Preprocessing & data exploration

Drop low-value columns

Begin by identifying columns that add little-to-no value for predictive modeling. These columns will be dropped.

The first, most obvious candidate to be dropped is PassengerID. No information was provided to us as to how these keys were derived. Therefore, the keys could have been completely random and may add false correlations or noise to our model.

The second candidate for removal is the passenger Name column. Normally, names can be used to derive missing values of gender, but the gender column holds no empty values. Thus, this column is of no use to us, unless we use it to engineer another name column.

The third candidate for removal will be the Ticket column, which represents the ticket serial ID. Much like PassengerID, information is not readily available as to how these ticket strings were derived. Advanced users may dig into historical documents to investigate how the travel agencies set up their ticket names, perform a clustering analysis, or bin the ticket values. Those techniques are out of the scope of this experiment.

The last candidate to be dropped will be Cabin, which is the cabin number where the passenger stayed. Although this column may hold value when binned, there are 147 missing values in this column (~21% of the data). Advanced users may cluster the cabins by letters, or can dig down into the grit of the actual RMS Titanic ship schematics to derive useful features such as cabin distance from hull breach or average elevation from sea level.

Define categorical variables

We must now define which values are non-continuous by casting them as categorical. Mathematical approaches for continuous and non-continuous values differ greatly. For example, if we graph the “Survived” column now, it will look funny because it would try to account for the range between “0” and “1”. However, being partially alive in this case would be absurd. Categorical values are looked at independently of one another as “choices” or “options” rather than as a numeric range.

For a quick (but not exhaustive) exercise to see if something should be categorical, simply ask, “Would a decimal interval for this value make sense?”

From this exercise, the columns that should be cast as categorical are: Survived, Pclass, Sex, and Embarked. The trickiest of these to determine might have been Pclass because it’s a numerical value that goes from 1 to 3. However, it does not really make sense to have a 2.5 Class between the second class and third class. Also, the relationship or “distance” between each interval of PClass is not explicit.

To cast these columns, drag in the “Edit Metadata” module. Specify the columns to be cast, then change the “Categorical” parameter to “Make categorical”.

Clean missing data in Azure ML

Most algorithms are unable to account for missing values and some treat it inconsistently from others. To address this, we must make sure our dataset contains no missing, “null,” or “NA” values. There are many ways to address missing values. We will cover three: replacement, exclusion, and deletion.

We used exclusion already when we made a conscious decision not to use “Cabin” attributes by dropping the column entirely.

Replacement is the most versatile and preferred method because it allows us to keep our data. It also minimizes collateral damage to other columns as a result of one cell’s bad behavior. In replacement, numerical values can easily be replaced with statistical values such as mean, median, or mode. The median is usually preferred for machine learning because it preserves the distribution of the data and is less affected by outliers. However, the median will skew and overload your frequencies, meaning it’ll mess with your bar graph but not your box plot.

We will cover deletion later in this section.

Now we can hunt for missing values. Drag in a “Summarize Data” module and connect it to your “Edit Metadata” module. Run the experiment and visualize the summary output. You will get a column summarizing the “missing value count” for each attribute. At this point, there are 177 missing values for “Age” and 2 missing values for “Embarked.”

Looking at the metadata of “Age” reveals that it is a “numeric” type. As such, we can easily replace all missing values of age with the median. In this case, each missing value will be replaced with “28.”

Embarked is a bit trickier since it is a categorical string. Usually, the holes in categorical columns can be filled with a placeholder value. In this case, there are only 2 missing values so it would not make much sense to add another categorical value to “Embarked” in the form of S, C, Q, or U (for unknown) just to accommodate 2 rows. We can stand to lose 0.2% of our data by simply dropping these rows. This is an example of deletion.

To clean missing values in Azure ML, use the “Clean Missing Data” module. This module will apply a single blanket operation to the selected features.  First, we start by having one “Clean Missing Data” module to replace all missing numeric instances with the median.  To select all the numeric columns, we select “Column Type” and “Numeric” under “Launch Column Selector” in the Properties of “Clean Missing Data.”

This will target only the “Age” column since it is the only numeric column with missing values. After the data goes through the module, there should only be 2 missing values left in the entire dataset, which is in “Embarked” column.  Then, we add another “Clean Missing Data” module, set it to drop the missing rows in order to remove the 2 missing values of “Embarked.”

Specify a response class

We must now directly tell Azure ML which attribute we want our algorithm to train to predict by casting that attribute as a “label.” Do this by dragging in a “Edit Metadata” module. Use the column selector to specify “Survived” and change the “Fields” parameter to “Labels.” A dataset can only have 1 label at a time for this to work. Our model is now ready for machine learning!

Partition and withhold data

It is extremely important to randomly partition your data prior to training an algorithm to test the validity and performance of your model. A predictive model is worthless to us if it can only accurately predict known values. Withhold data represents data that the model never saw when it was training its algorithm. This will allow you to score the performance of your model later to evaluate how well the model can predict future or unknown values.

Drag in a “Split Data” module. It is usually industry practice to set a 70/30 split. To do this, set “fraction of rows in the first output dataset” to be 0.7. 70% of the data will be randomly shuffled into the left output node, while the remaining 30% will be shuffled into the right output node.

Select an algorithm

First, we must identify what kind of machine learning problem this is: classification, regression, clustering, etc. Since the response class is a categorical value, or “0” or “1”, for survived or deceased, we can tell that it is a classification problem. Specifically, we can tell that it is a two-class, or binary, classification problem because there are only two possible results: survived or deceased. Luckily, Azure ML ships with many two-class classification algorithms. Without going into algorithm-specific implementations, this problem lends itself well to decision forest and decision tree because the predictor classes are both numeric and categorical. Pick one algorithm (any two-class algorithm will work).

Train your model

Drag in a “Train Model” module and connect your algorithm to it. Connect your training data (the 70%) to the right input of the “Train Model” module. To score the model, drag in a “Score Model” module. Connect the “Train Model” to the left input node of the “Score Model,” and the 30% withhold data to the right input node of the “Score Model.” Finally, to evaluate the performance of model, drag in an “Evaluate Model” module and connect its left input to the output of the “Score Model.”

Run your model

Evaluate your model

If you visualize your “Evaluate Model” module after running your model, you see a staggering number of metrics. Each machine learning problem will have its own unique goals, thus having different priorities when evaluating “good” or “bad” performance. As a result, each problem will also optimize different metrics.

For our experiment, we chose to maximize the RoC AuC because this is a low-risk situation where the outcomes of false negatives or false positives do not have different weights.

RoC AuCs will vary slightly because of the randomized split. The default parameters of our two-class boosted decision tree yielded a RoC AuC of 0.832. This is a fair-performing model. By fine-tuning the parameters, we can further increase the performance of the model.

Which metric to optimize?

1. RoC AuC: Overall Performance
2. Precision: Relevance
3. Recall: Thoroughness
4. Accuracy: Correctness

Beginner’s guide to RoC AuC

• o.9~1 = Suspiciously Good
• 0.8~0.9 = Fair
• 0.7~0.8 = Decent Model
• 0.5~0.6 = Worthless Model

Compare your model

How would our model shape up against another algorithm? Let’s find out. Drag in a “Two-Class Decision Forest” module. Copy and paste your “Train Model” module and your “Score Model” module. Reroute the input of the newly-created “Train Model” module to the decision forest. Attach the output of the newly-created “Score Model” module to the right input node of the “Evaluate Model” module. Now we can compare the performance of two machine learning models that were trained separately.

Both models performed fairly (~0.81 RoC AuC each). The boosted decision tree got a slightly higher RoC AuC overall, but the two models were close enough to be considered tied in terms of performance. As a tiebreaker, we can look at other metrics such as accuracy, precision, and recall. Using those metrics, we found that the boosted decision tree had lower accuracy, precision, and recall when compared to the two-class decision forest. If we were to select a winning model right now, it would probably be the two-class decision forest.

Other video tutorials

You can watch this series of videos to dive deeper into Azure Machine Learning:

• What is Azure Machine Learning?
• Subscriptions and Workspaces
• Import and Export Data, Modules, and Experiments
• Data Exploration
• Renaming Columns and Replicating Data
• Joining Datasets
• Dropping and Selecting Columns
• Cleaning and Summarizing Data
• Splitting and Categorical Casting
• Building your Machine Learning Model

As machine learning applications become accessible, organizations will adopt ML principles in their demand planning process, leading to better management.

What is demand planning?

It is expensive to build products that don’t sell well – Businesses incur warehouse costs to store excess inventory, and they incur expenses each time they move or repackage inventory. Finally, they frequently need to dispose of old inventory as it becomes outdated or expired. With all these costs combined, it costs businesses an average of $20 per year to manage $100 of inventory.

Because of the high cost of holding excess inventory, businesses develop forecasts of future demand through a process known as Demand Planning.

Demand Planning is the process of forecasting future demand for a product so that the supply chain has sufficient inventory to satisfy customer demand.

By knowing the likely sales of future products, the business’s supply chain can produce just enough products to satisfy customer demand, while at the same time not creating vast amounts of unsold inventory.

Complexities of demand planning: Example with Breakfast Cereal

The demand planning process is very hard to do well. Most media and large companies do quite well at forecasting total sales, but it gets quite hard to forecast sales at a product-specific level. Consider the sales of a major producer of breakfast cereal. In total, sales of all cereal varieties combined are pretty stable. However, when you analyze the sales of specific varieties, it is clear that some varieties are more predictable than others. Sales of both corn flakes and cocoa-flavored cereal are historically pretty stable.

These two varieties are likely to be forecasted based on a rolling average of sales in prior months, or a simple time series model.

However, sales of specialty cereals are volatile every month. Sales of Specialty #1 increased dramatically between Month 1 and Month 4, before declining again in Month 5. If you had not correctly anticipated the drop-off of sales in Months 5 and 6, it is quite possible you would have over-produced inventory.

In this quick illustration, I reveal the challenges of forecasting just four different product types. But, a large cereal manufacturer will likely have dozens of varieties – They may produce gluten-free, reduced sugar, sugar-free, various flavors, different package sizes, and unique product variants for each market or country. This quickly becomes a massive exercise to predict sales of each product!

Regression Analysis: A Static Comparison

To predict sales of highly-volatile products, skilled demand planners frequently rely on correlation with outside factors. While historical sales of any given product may appear quite random, the demand planner realizes that much of the volatility can be explained by creating a regression model against other data.

Continuing with the example of the sales of cereal, let’s assume Cereal Specialty #1 carries the theme of a recently released children’s movie. Thus, it is quite conceivable that sales of this cereal variety will be correlated with this movie’s performance at the box office. Once the children see the movie, they (or their parents) are likely to see the cereal on their next trip to the store.  In this case, the demand planners at the cereal business may model box office sales for this movie as a leading indicator for this cereal variety, which carries the theme of the movie.

Machine Learning: Adapting to a Changing World

Up to this point, I’ve referenced both time-series and simple regression models as traditional techniques that demand planners employ to predict future product demand. Both of these techniques are essentially static models. That is, the forecast models utilize a defined set of inputs (either historical sales or defined market data).

Today, a vast amount of useful data is readily available – All types of data, including industry trends, economic data, and even weather forecasts and social network feeds, could be used to improve the predictive power of demand planning. For instance, continuing with the example of cereal sales – a recent study revealing previously unknown health benefits of oat bran may impact sales of oat bran cereal in the coming weeks.

Alternatively, a recent surge in consumer-packaged coffee may drive slightly increased sales of cereals that are paired well with coffee. Social signals on a movie release will likely impact sales of cereals that are themed with the movie.

Each of these factors, in isolation, may contribute to only slight increases in forecast accuracy. However, when several of these factors are combined, they may generate a dramatic improvement. While this in theory is great, the reality is that it would be cost-prohibitive to employ enough analysts to run statistical analyses against all possible factors.

Because of this, demand planning is an ideal application for machine learning. Through the use of sophisticated algorithms, machine learning applications can generate highly-predictive forecasts from enormous amounts of disparate data. The user typically inputs a training dataset, which may contain any number of source data that may be correlated with product sales.  The software then identifies which data series are correlated with product sales, and generates an algorithm for the forecast model.

The key benefit of the machine learning process is that it is dynamic, meaning that the user can add datasets at any time, and allow the software to determine if the dataset is a good predictor of sales for any product category. As a simple example: Sales of umbrellas may be correlated with rainy days.

Upon further analysis, the user may want to know if nearby sporting events have any impact on umbrella sales. It would take a long time to run a regression analysis against every rainy day and sporting event. Instead, a demand planner could create a training dataset with historical sales data, weather conditions, and sporting events. With machine learning, he can quickly see which factors impact umbrella sales for a particular region.

Rapid Adoption of Machine Learning

Most large universities now offer programs specifically tailored to data analytics professions. Similarly, organizations like Data Science Dojo provide training specifically geared to the adoption of data science.  Accordingly, it will continue to get easier to find people with both an interest and experience in the machine learning profession.

In addition, software tools designed for machine learning are becoming accessible to the general public.  For instance, Amazon Web Services now offers a forecasting tool that allows virtually anyone to build and deploy a demand forecast using machine learning.

Conclusion

Machine Learning Will Transform Demand Planning

As machine learning applications become more accessible, we will see more organizations adopt machine learning principles in their demand planning process. Instead of relying on the decades-old strategy of using time-series analysis or simple regression, supply chains will become more optimized. Today, businesses in most industries maintain  60 to 90 days of inventory. Once machine learning becomes more widespread, we will see businesses routinely manage with less than 30 days of inventory on hand.

Check out some of the best machine learning courses.

Angela Baltes completed Data Science Dojo’s bootcamp program at the University of New Mexico. Here’s her reflection on the course and the bootcamp review.

The opportunity to participate in Data Science Dojo’s: A Hands-on Introduction to Data Science bootcamp was a simple decision, as I have been a consumer of bootcamps for several years and have found my success varies with them.

In my prior self-paced learning, I found that there were concepts that I simply did not understand well, or perhaps were not explicitly stated in whatever course I was taking.

I wanted to experience an immersive in-person bootcamp with the hopes that practical examples and in-person interactions would be helpful in understanding and retaining the material. Not to mention, I was able to network with others who are interested in this field.

Data Science  is taught by Raja Iqbal, CEO and Chief Data Scientist. He is a talented presenter, and I appreciated his style of teaching the material. He was accompanied by Arham Akheel, who assisted Raja in helping students and provided us with machine learning demonstrations.

This combination was very complimentary to one another and worked well. Please check out Data Science Dojo’s website and check their schedule for they may be coming to a city near you!

5-day Data Science Bootcamp

The bootcamp was offered in Albuquerque, New Mexico for 3 days instead of the prior 5-day bootcamp. From what I understand, we were the first cohort to try this format.

Day 1

On this first day, we spent some time looking into data exploration, and how to approach data problems. We discussed things as a group, and I enjoyed the energy from class.

We discussed that a model is only as good as the data provided to it-garbage in, garbage out. Data is the new oil and is the most valuable asset a company can have, however, we, as data scientists, need to tap into that resource by refining it and getting the most value from it.

One thing that I have personally struggled with is that this course was extremely helpful for learning how to ask the right questions and evaluate business impact. It is our job to ask questions.

Many times, in the past, I was given a task, and I simply began to hammer away without questions asked. In data science, feature engineering and data exploration are the most important tasks, as these activities help to further define and evaluate if this is a worthy endeavor for a company.

Day 2

On this day, we began to delve into machine learning algorithms, more specifically, supervised learning. I found this valuable, as I myself, have the most experience with and understanding of supervised learning. We stressed again before building a model to ask, “What is the intended use of this model?”  as that would be pertinent information in determining what features and format to provide the model to the stakeholders’ that will use it.

We analyzed the Titanic dataset in detail and discussed what features to include in our decision tree model. We also discussed entropy, stopping criteria, and splitting. Our homework assignment was to submit our Titanic model to our leaderboard. I did not place very high, lol.

Day 3

On the last day of the bootcamp, we discussed the pitfalls in machine learning, such as overfitting and underfitting, and understood the bias/variance tradeoff. I have read about this topic to the point of nausea in other settings, but this truly helped me to understand it. Seeing practical examples helped me put this in context.

What was interesting and new to me was discussing how to properly evaluate a model, as it is not always about the accuracy-sometimes (depending upon the problem and domain), it is about the precision or recall! We then spent a great deal of time on hyperparameter tuning, and then how to deploy our machine learning model as a web service, which was way too cool.

What I’ve Learned

I did not completely understand how to tune hyperparameters and how to properly evaluate the performance of a model before the bootcamp. Now I understand why this is necessary and how to carry out this task. We bridged the gap between data science and business value in this course, and that was the foundation going forward.

What I learned is that it is not always about the accuracy of a model, and to align the business needs with precision or recall depending upon the domain and problem one is looking to solve.

I have learned why it is important for the data scientist to ask questions, and not just questions in general, but the right questions, and how the most important tasks before building a model are data exploration, data discovery, and feature engineering. We need to understand the business impact and how this model will add value.

For me, this was paramount. Too many times do we focus on wanting cool models to say we are involved in machine learning rather than focusing on the business need.

I have learned how to use Microsoft tools to build and deploy a model as a web service. I found the ease and simplicity of this to be amazing and something I would like to continue to explore.

The Pros

• The in-person class setting was helpful in order to understand and connect to the topics at hand. For those who have taken online bootcamps with varying success, you may also appreciate being able to interact with the instructor and other students.
• The breadth of material covered was impressive. I appreciated that we covered the most important topics in machine learning and addressed common mistakes. We dedicated some of the day to hyperparameter tuning when a model is not performing optimally.
• We addressed the proper mindset to have for data analysis. How to ask the right questions, and not be afraid to ask questions!
• Raja and Arham have great chemistry as team members and are fantastic instructors.

The Cons

• The condensed format was rather overwhelming. This material isn’t truly suited for a 3-day setting. We really only scratched the surface. This cannot be truly helped, but it was worth mentioning.
• This course is not for those who are new to programming and/or data science. Although we did use Microsoft Azure for machine learning, there is an assumption that the student has some familiarity with programming and data science concepts. You will likely get more out of this course if you have some prior knowledge.

Summing Up the Bootcamp Review

I highly recommend this bootcamp for those who would like to increase their knowledge in data science. This experience was valuable for me so that I can bridge the gap between theory and implementation. From this point on, more learning will be required, but this gave me the boost in the right direction. Cheers!

This review was originally published on Angela Baltes’ personal blog.