Churn Prediction – Project Details

CompanyJio Telecom
RoleML Analyst / Data Scientist
Duration4 Weeks
Skill LevelAdvanced
Tools UsedPython, Pandas, Scikit-learn, Tableau
DeliverablesPredictive ML model, Feature engineering pipeline, Churn analysis dashboard
Objective

Predict customer churn in telecom by building a classification model using behavioral and usage data from Jio customers. The model identifies high-risk users based on patterns like call drop rates, data usage, plan downgrades, and inactivity periods. Feature engineering techniques like encoding, scaling, and binning are applied to prepare the dataset. The final model integrates with a real-time dashboard enabling customer retention teams to trigger proactive engagement actions.

Key Challenges
  • Class imbalance in churned vs active users
  • Missing values in usage data
  • Identifying non-obvious churn indicators
Features
  • Churn probability score per customer
  • Usage heatmaps and behavior metrics
  • Real-time dashboard with filters
Outcome

Improved retention by 18% through targeted offers to churn-risk users. Model achieved 86% accuracy on validation set.