Telecom Customer Churn Prediction

Project Goal

The objective of this project is to build a machine learning model that predicts customer attrition (churn) for a telecommunications provider. By identifying at-risk customers, the business can implement targeted retention strategies to reduce revenue loss.

The Dataset

The analysis is performed on a dataset containing 7,043 customer records with 21 features, including:

• Demographics: Gender, Senior Citizen status, Partners, and Dependents.
• Services: Phone service, Multiple lines, Internet (DSL/Fiber optic), and Security/Support add-ons.
• Account Info: Tenure, Contract type (Month-to-month, One year, Two year), Payment method, and Charges.

Technical Workflow

1. Data Cleaning: - Handled non-numeric values and nulls in the TotalCharges column.
   • Performed data type conversion to ensure all features were ready for modeling.
2. Exploratory Data Analysis (EDA): - Visualized churn distribution across different categories.
   • Identified that customers with month-to-month contracts and fiber optic internet had higher churn rates.
3. Feature Engineering & Scaling:
   • Encoded categorical variables for machine learning compatibility.
   • Scaled numerical features using StandardScaler.
4. Machine Learning Modeling:
   • Trained a Logistic Regression model for interpretability.
   • Trained a Random Forest Classifier to maximize predictive power.
5. Evaluation:
   • Utilized a Confusion Matrix to track True Positives/Negatives.
   • Calculated the ROC-AUC score to measure the model's ability to distinguish between churners and non-churners.

Libraries Used

• Pandas & NumPy: Data manipulation and cleaning.
• Matplotlib & Seaborn: Statistical data visualization.
• Scikit-Learn: Model building, scaling, and evaluation metrics.