Customer Churn Analysis and Prediction

This comprehensive repository provides tools and insights for understanding and predicting customer churn in the context of an internet service provider. The analysis encompasses both exploratory data analysis and predictive modeling to equip you with a holistic approach to customer retention.

Table of Contents

1. Data Exploration:

   • Investigate churn rates across various customer segments:
     • Internet service types
     • Customer partnerships
     • Contract lengths
     • Payment methods
     • Gender

2. Analysis:

   • Uncover potential reasons and factors influencing churn variations across different segments.

3. Calculations:

   • Compute precise churn percentages for each identified segment.

4. Findings & Recommendations:

   • Provide actionable insights and recommendations to address high priority areas for reducing customer turnover.

Key Results & Observations

The analysis has revealed several key insights:

• Fiber Optic Customers: Exhibit the highest churn rate at 42%, indicating potential issues with reliability, speed, or cost.

• Partnerships and Longer Contracts: Strongly correlate with improved retention, suggesting loyalty incentives as effective churn reduction strategies.

• Payment Methods: Automatic credit card payments are associated with the lowest churn, while electronic checks show a concerning 45% churn rate.

• Gender: Does not appear to be a standalone indicator for likelihood to churn.

Predictive Modeling Insights

For those seeking a predictive approach, consider the following machine learning insights:

• Oversampling with SMOTE: Improved model recall by 15%, addressing imbalances in the dataset.

• XGBoost Algorithm: Achieved the highest precision after hyperparameter optimization, showcasing its effectiveness for this task.

• Top Predictors: Identified contract type and payment method as crucial predictors in forecasting churn.

Usage & Extensions

This repository serves as a versatile template for churn analysis and prediction. Tailor it to your specific needs:

• Further Exploration: Expand on service performance, usage behaviors, and additional metrics for a more nuanced understanding.

• Machine Learning Models: Build on the predictive aspect by implementing models to forecast individual customer churn risk.

• Revenue Analysis: Extend the analysis to include revenue metrics for different customer cohorts.

• Intervention Opportunities: Identify specific opportunities for targeted interventions to mitigate churn.

License

This project is licensed under the MIT License.