Customer Churn Prediction – Telecom Sector

Organization & Duration

Cognizant Case Study | 7 Days

Techniques & Tools

• Logistic Regression, Random Forest, Gradient Boosting, Hyperparameter Tuning
• Python, scikit‑learn, Power BI

Key Outcomes

• Achieved AUC = 0.86 and 81% accuracy.
• Identified key churn predictors such as contract type and tenure.
• Delivered retention strategy recommendations and scoring models.

Insights

• Random Forest and Gradient Boosting achieved AUC = 0.86 and 81% accuracy → strong churn classification performance.
• Contract type, tenure, and monthly charges emerged as the top churn predictors.
• Developed a churn scoring model and retention strategy focused on high‑risk customer segments.

Business Impact

This churn prediction system empowers telecom providers to proactively retain customers. By identifying high‑risk segments and modelling churn probability, the project supports targeted retention campaigns, optimized contract offerings, and improved customer satisfaction.

View on GitHub ← Back to Home