A case study on predicting customer churn using machine learning.
Customer churn is the loss of customers. It's a key success metric for many businesses. It's also important because from an economic perspective, it costs much less to keep customers than to get new ones.
Historically customer churn reduction has been addressed with slow iterations of A/B testing on web page changes. My approach here uses machine learning to uncover the components driving the churn so the business can strategically understand causes (and opportunities) and proactively address them.
To see an explanation of the problem and a step by step machine learning solution, click on the file "Predict_Customer_Churn_Case_Study.ipynb". This opens an ipython (Jupyter) notebook with code, narrative, and data visualizations.
This approach uses statistical analysis to first understand relationships in the data. Then I apply machine learning using an ensemble bagging method (XGBoost), do some feature engineering, and hyperparameter tuning with GridSearchCV.