This project focuses on predicting telecom customer churn using machine learning techniques. Customer churn prediction helps businesses identify customers likely to discontinue services, enabling proactive retention strategies.
The project includes data preprocessing, exploratory data analysis (EDA), feature engineering, machine learning model building, performance evaluation, and model comparison.
- Analyze customer behavior and churn patterns
- Identify major factors influencing churn
- Build predictive machine learning models
- Compare multiple classification algorithms
- Support customer retention decision-making
Dataset: Telco Customer Churn Dataset
The dataset contains telecom customer details such as:
- demographic information
- subscription details
- billing information
- service usage
- churn labels
Key features include:
- Gender
- Senior Citizen
- Partner
- Dependents
- Tenure
- Monthly Charges
- Total Charges
- Internet Service
- Contract Type
- Payment Method
- Tech Support
- Online Security
- Streaming Services
Target Variable:
- Churn Label
- Python
- Pandas
- NumPy
- Matplotlib
- Seaborn
- Scikit-learn
- Joblib
- Jupyter Notebook
Used as a baseline classification model for churn prediction.
Used for improved prediction performance and feature importance analysis.
| Model | Accuracy |
|---|---|
| Logistic Regression | ~80% |
| Random Forest | ~79% |
- Customer churn distribution
- Confusion matrix
- Feature importance analysis
- Model accuracy comparison
- Customers with higher monthly charges are more likely to churn.
- Long-term customers show significantly lower churn rates.
- Month-to-month contract users exhibit higher churn probability.
- Billing and contract-related features strongly influence churn.
- Logistic Regression and Random Forest delivered strong predictive performance.
Telecom-Customer-Churn-Prediction-Retention-Analysis/
│
├── customer_churn_prediction.ipynb
├── Telco_customer_churn.xlsx
├── customer_churn_model.pkl
├── requirements.txt
└── README.md