Customer Churn Prediction

Customer churn is a critical issue in the telecommunications industry, impacting profitability and customer retention. This project uses machine learning to predict churn, enabling businesses to take proactive measures and improve customer satisfaction.

Key Outcomes

Achieved 81.7% accuracy using Logistic Regression, with a weighted F1 score of 0.81.
Random Forest model was tested but achieved a slightly lower accuracy of 80%.
Top predictor of churn: Month-to-Month contract type, emphasizing the importance of subscription models in churn behavior.

Project Highlights

Data Analysis: Conducted in-depth analysis of customer demographics, service details, and account information to identify trends and correlations.
Machine Learning Models:
- Logistic Regression achieved a balanced performance with high precision and recall.
- Random Forest model tested as a benchmark for comparison.
Deployment: Designed a user-friendly Flask application to allow users to upload datasets and get predictions in real-time.

Visual Results

Logistic Regression Performance:

Logistic Regression Classification Report

Key Insights:

Customers with Month-to-Month contracts are at the highest risk of churn.
Identified critical churn predictors, including tenure length and internet service type.

Business Impact

This solution enables businesses to:

Proactively retain customers by identifying at-risk groups early.
Personalize marketing strategies based on churn predictors.
Enhance profitability by reducing churn rates and increasing customer lifetime value.

Tools and Technologies

Programming: Python, Flask
Libraries: Pandas, Scikit-learn, Matplotlib
Deployment: Flask for real-time predictions
Dataset: ## Dataset

The dataset used in this project is available for download:
Telco Customer Churn Dataset

Call to Action

Interested in how I can bring similar insights to your organization?
Let’s connect on LinkedIn or explore the GitHub repository.