Customer Churn Prediction

01 · The problem

A subscription business needed to spot high-risk customers early enough to intervene.

02 · Approach

1. 1

   Ran EDA on 10,000 customer records.

2. 2

   Resolved class imbalance with SMOTE.

3. 3

   Trained and compared Random Forest, XGBoost, and Logistic Regression in Scikit-learn.

03 · Outcome

Random Forest was the best model. Top churn drivers: days since last interaction, monthly bill charges, and account tenure.