Churn Analysis Predictive Modeling

Introduction

The key to most marketing and product campaigns is to attract new customers and at the same time reduce attrition rates(churn). Churn can be triggered for different reasons. Therefore studying churn behavior and utilizing insights drawn from such studies can be very beneficial inorder for companies to grow their business and revenue. This is a repository that contains churn analysis and predictive modeling techniques that I learnt when I was completing my research work during my study in MS Computer Science at Santa Clara University.

Pipeline Steps

A basic machine learning pipline was built and perfromance of different model types were compared.

Step 1: Understanding Problem Statement

Step 2: Data Collection

Step 3: Exploratory Data Analysis (EDA)

Conducted exploratory data analysis through visualizations on Tableau to answer questions and derive actionable insights

Step 4: Feature Engineering

• Handling imbalanced data
• Applying label encoding for binary features
• Converting categorical variable into dummy variables

Step 5: Train/Test Split

Step 6: Model Evaluation Metrics Definition : Confusion matrix

Step 7: Model Selection, Training, Prediction and Assessment : SVM, Random Forest, Logistic Regression, and XGBoost

Step 8: Hyperparameter Tuning/Model Improvement : Logistic Regression, SVM

References

https://towardsdatascience.com/machine-learning-case-study-telco-customer-churn-prediction-bc4be03c9e1d