📈 Customer Churn Prediction & Retention Strategy

📌 Overview

This project applies machine learning to customer data from a fictional telecommunications company to predict customer churn. The core objective is to move beyond simply identifying churn by developing a robust predictive model. This model enables the company to proactively reach out to at-risk customers with targeted offers designed to persuade them to stay.

The analysis identifies the main factors driving customer attrition to guide smarter, data-driven retention decisions.

---

🎯 Objectives

• Data Preprocessing: Clean and prepare the Telco customer dataset, including handling missing values (11 entries in TotalCharges) and converting data types.
• Exploratory Data Analysis (EDA): Explore the key factors (e.g., contract type, tenure, services) that have the biggest impact on a customer's likelihood to churn.
• Model Building: Build and compare different classification machine learning models to find the best performer for prediction.
• Recommendation: Select the most suitable model for real-world use and present clear, actionable insights to guide concrete business actions.

---

📊 Methods

• Data Source: The project utilizes the Telco Customer Churn dataset from Kaggle, with minor modifications.
• Data Cleaning: Missing values were removed (11 rows from over 5,000 entries). Numeric columns (MonthlyCharges, TotalCharges) were converted to the correct data type.
• Analysis: Visualizations (e.g., boxplots) were used to analyze feature dependencies, specifically comparing churn against key variables like tenure, MonthlyCharges, and TotalCharges.
• Model Selection: Multiple classification models were considered, with Logistic Regression ultimately selected as the final recommended model for its balance of performance and interpretability.
• Libraries Used: numpy, pandas, matplotlib.pyplot, and seaborn.

---

🔑 Key Results

The exploratory analysis revealed distinct characteristics of customers likely to churn:

• Risk Factors: Customers with lower tenure, higher monthly charges, and lower total charges are significantly more likely to churn.
• High-Risk Segments: The highest churn rates are associated with:
  • Contract Type: Month-to-Month.
  • Internet Service: Fiber Optic.
  • Payment Method: Electronic Check.
• Final Model: Logistic Regression was chosen as the most robust and useful model for prediction.

---

💡 Implications

• Retention Focus: The insights suggest that providing better onboarding, early engagement, or promotional contracts for new customers may be crucial to reducing churn.
• Business Action: The predictive model can be deployed to target interventions. A recommended decision threshold between 0.1 and 0.3 is suggested, which can be adjusted based on the company's budget for retention offers and tolerance for false positives.

---

📁 Repository Structure

├── data/
│   └── churn_data_set_sample.csv # Input dataset (a sample or version of the data)
├── notebook/
│   └── customer_churn_analysis.ipynb # The main analysis and modeling notebook
├── README.md
└── LICENCE

---

🚀 Getting Started

1. Clone Repository

git clone <repository-url>
cd <repository-name>

2. Run Jupyter Notebook

Open the main notebook to view the complete analysis, cleaning steps, and model evaluation.

---

📚 References

Telco Customer Churn on Kaggle

Scikit-learn Documentation

---

📝 License

Distributed under the MIT License. See LICENSE for details.

---

✉️ Contact

For more info and collaboration requests: connect with me on LinkedIn.