Project 2: Customer Churn Prediction

Predict which customers will leave a service (churn prediction).

Starter code: Run starter.py in this folder after downloading the dataset (see Dataset below).

Difficulty

Intermediate

Time Estimate

4-5 days

Skills You'll Practice

• Classification
• Feature Engineering
• Imbalanced Data Handling
• Business Metrics
• Model Interpretation

Learning Objectives

By completing this project, you will learn to:

• Handle imbalanced datasets
• Apply business-focused metrics
• Engineer features from customer data
• Use ensemble methods
• Interpret model results for business

Dataset

Telco Customer Churn Dataset

• Kaggle Telco Customer Churn
• Customer information and churn status
• Features: demographics, services, charges, etc.

Project Steps

Step 1: Load and Explore Data

• Load dataset
• Analyze churn rate
• Explore feature distributions
• Identify missing values
• Check for class imbalance

Step 2: Data Preprocessing

• Handle missing values
• Encode categorical variables
• Scale numerical features
• Create train/test split

Step 3: Feature Engineering

• Create new features (e.g., tenure groups, charge ratios)
• Handle categorical variables
• Feature selection
• Correlation analysis

Step 4: Handle Class Imbalance

• Use SMOTE for oversampling
• Adjust class weights
• Try different sampling strategies
• Compare approaches

Step 5: Model Training

• Train multiple models:
  • Logistic Regression
  • Random Forest
  • XGBoost
  • Ensemble methods
• Use cross-validation

Step 6: Model Evaluation

• Calculate accuracy, precision, recall, F1
• Focus on recall (finding churners)
• Calculate business metrics (cost of churn)
• Feature importance analysis

Step 7: Business Insights

• Identify key churn factors
• Recommend retention strategies
• Calculate ROI of interventions
• Create actionable insights

Expected Deliverables

1. Jupyter Notebook with complete analysis
2. Model with good recall for churners
3. Business Report with insights and recommendations
4. Dashboard (optional) showing key metrics

Evaluation Metrics

• Recall: Most important (find actual churners)
• Precision: Minimize false positives
• F1-Score: Balance
• Business Cost: Calculate cost of churn vs retention

Key Features to Explore

• Tenure (how long customer has been with company)
• Monthly charges
• Services used
• Contract type
• Payment method

Tips

• Focus on recall (finding churners is more important)
• Calculate cost of false negatives vs false positives
• Feature engineering is crucial
• Try ensemble methods
• Interpret results for business stakeholders

Resources

• Kaggle Telco Churn Dataset
• Handling Imbalanced Data
• Business Metrics for ML

Extensions

• Build retention campaign recommendations
• Create real-time churn prediction system
• A/B test retention strategies
• Deploy model as API

Next Steps

After completing this project:

• Try other churn datasets
• Experiment with advanced techniques
• Move to Project 3: Movie Recommendation