Customer Churn Analytics using Microsoft R Open
7 likes2,284 views
The document summarizes a presentation on using Microsoft R Open for customer churn analytics. It discusses using machine learning algorithms like logistic regression, support vector machines, and random forests to predict customer churn. It compares the performance of these models on a telecom customer dataset using metrics like confusion matrices and ROC curves. The presentation demonstrates building a churn prediction model in Microsoft R Open and R Tools for Visual Studio.
Customer Churn Analytics using Microsoft R Open
- 1. Customer Churn Analytics using Microsoft R Open Malaysia R User Group Meet Up 16th February 2017 Poo Kuan Hoong https://github.com/kuanhoong/churn-r
- 2. Disclaimer: The views and opinions expressed in this slides are those of the author and do not necessarily reflect the official policy or position of Nielsen Malaysia. Examples of analysis performed within this slides are only examples. They should not be utilized in real-world analytic products as they are based only on very limited and dated open source information. Assumptions made within the analysis are not reflective of the position of Nielsen Malaysia.
- 3. Agenda • Introduction • Customer Churn Analytics • Machine Learning Framework • Microsoft R Open and Visual Studio • Model Performance Comparison • Demo
- 4. Malaysia R User Group (MyRUG) • The Malaysia R User Group (MyRUG) was formed on June 2016. • It is a diverse group that come together to discuss anything related to the R programming language. • The main aim of MyRUG is to provide members ranging from beginners to R professionals and experts to share and learn about R programming and gain competency as well as share new ideas or knowledge.
- 8. Introduction • Customer churn can be defined simply as the rate at which a company is losing its customers • Imagine the business as a bucket with holes, the water flowing from the top is the growth rate, while the holes at the bottom is churn • While a certain level of churn is unavoidable, it is important to keep it under control, as high churn rate can potentially kill your business
- 10. Churn analytics • Predicting who will switch mobile operator
- 11. Customer churn - who do customers change operators? • The top 3 reasons why subscribers change providers: • They want a new handset • They believe they pay too much for calls/data • Providers do not offer additional loyalty benefits
- 12. Data Collection Data Preprocessing Attributes selection • Attribute 1 • Attribute 2 • Attribute 3 Algorithm Training Model Score Model Apply Data /Test Data Predicting Output Initialization Step Learn Step Apply Step Machine Learning Framework
- 13. Correlation Matrix • correlation matrix, which is used to investigate the dependence between multiple variables at the same time.
- 14. Microsoft R Open • Microsoft R Open, formerly known as Revolution R Open (RRO), is the enhanced distribution of R from Microsoft Corporation. • It is a complete open source platform for statistical analysis and data science. Key enhancement • Multi-threaded math libraries that brings multi-threaded computations to R. • A high-performance default CRAN repository that provide a consistent and static set of packages to all Microsoft R Open users. • The checkpoint package that make it easy to share R code and replicate results using specific R package versions.
- 15. R Tools for Visual Studio • Turn Visual Studio into a powerful R development environment • Download R Tools for Visual Studio
- 16. R Tools for Visual Studio • Visual Studio IDE • Intellisense • Enhanced multi-threaded math libs, cluster scale computing, and a high performance CRAN repo with checkpoint capabilities. • Learn more about R Tools from here: https://microsoft.github.io/RTVS-docs/
- 17. Data Collection Data Preprocessing Attributes selection • Attribute 1 • Attribute 2 • Attribute 3 Algorithm Training Model Score Model Apply Data /Test Data Predicting Output Initialization Step Learn Step Apply Step Machine Learning Framework
- 18. Data Preprocessing • Assign missing values as zero • Detect outliers • Remove unwanted variables • Recode variables • Convert categorical variables
- 19. Data Collection Data Preprocessing Attributes selection • Attribute 1 • Attribute 2 • Attribute 3 Algorithm Training Model Score Model Apply Data /Test Data Predicting Output Initialization Step Learn Step Apply Step Machine Learning Framework
- 20. Features selection • The process of selecting a subset of relevant features (variables, predictors) for use in model construction. • Feature selection techniques are used for three reasons: • simplification of models to make them easier to interpret by researchers/users, • shorter training times, • enhanced generalization by reducing overfitting
- 22. Models Performance Comparison • Logistic Regression • is a regression model where the dependent variable (DV) is categorical. • Support Vector Machine • SVM is a supervised learning model with associated learning algorithms that analyze data used for classification and regression analysis. • RandomForest • is an ensemble learning method for classification, regression and other tasks, that operate by constructing a multitude of decision trees at training time and outputting the class that is the mode of the classes (classification) or mean prediction (regression) of the individual trees.
- 23. Data Collection Data Preprocessing Attributes selection • Attribute 1 • Attribute 2 • Attribute 3 Algorithm Training Model Score Model Apply Data /Test Data Predicting Output Initialization Step Learn Step Apply Step Machine Learning Framework
- 24. Training Model and Algorithm • Split the data set into 80:20 using library(caret) • Apply the algorithms: GLM, SVM and RF
- 25. Data Collection Data Preprocessing Attributes selection • Attribute 1 • Attribute 2 • Attribute 3 Algorithm Training Model Score Model Apply Data /Test Data Predicting Output Initialization Step Learn Step Apply Step Machine Learning Framework
- 26. Score Model • Confusion Matrix: a table that is often used to describe the performance of a classification model (or "classifier") on a set of test data for which the true values are known. • true positives (TP): These are cases in which we predicted yes (they have the disease), and they do churn. • true negatives (TN): We predicted no, and they don't churn. • false positives (FP): We predicted yes, but they don't actually churn. (Also known as a "Type I error.") • false negatives (FN): We predicted no, but they actually do churn. (Also known as a "Type II error.")
- 27. Confusion Matrix: Generalized Linear Model (glm) n=1407 Predicted: NO Predicted: YES Actual: NO TN = 919 (0.653) FP = 115 (0.082) 1034 Actual: YES FN = 167 (0.119) TP = 206 (0.146) 373 1086 321
- 28. Confusion Matrix: Support Vector Machine (SVM) n=1407 Predicted: NO Predicted: YES Actual: NO TN= 929 (0.660) FP= 105 (0.075) 1034 Actual: YES FN= 183 (0.130) TP= 190 (0.135) 373 1112 295
- 29. Confusion Matrix: RandomForest n=1407 Predicted: NO Predicted: YES Actual: NO TN= 939 (0.667) FP= 95 (0.068) 1034 Actual: YES FN= 181 (0.129) TP= 192 (0.136) 373 1120 287
- 30. Receiver Operating Characteristic (ROC) curve • ROC curve is a graphical plot that illustrates the performance of a binary classifier system as its discrimination threshold is varied. The curve is created by plotting the true positive rate (TPR) against the false positive rate (FPR) at various threshold settings.
- 31. Models comparison • ROC illustrates the performance of a binary classifier system as its discrimination threshold is varied.
- 32. Microsoft R Open vs R
- 33. Data Collection Data Preprocessing Attributes selection • Attribute 1 • Attribute 2 • Attribute 3 Algorithm Training Model Score Model Apply Data /Test Data Predicting Output Initialization Step Learn Step Apply Step Machine Learning Framework
- 34. Predict test data • Based on the training model, select the best model to be used for test data prediction
- 36. DEMO