Customer Lifetime Value (CLV) Prediction

Customer Lifetime Value prediction is crucial for any business that wants to:

• Focus marketing efforts on high-value customers
• Optimize resource allocation
• Make data-driven decisions about customer acquisition costs
• Build targeted retention strategies This project uses machine learning to predict CLV and then segments customers into high, medium, and low value groups. No more guesswork – just solid predictions backed by data.

Project Structure

• ├── data/processed/ # Clean, processed datasets ready for modeling
• ├── models/ # Trained model files
• ├── outputs/ # Generated predictions and visualizations
• ├── results/ # Analysis results and performance metrics
• ├── 1_data_processing.py # Data cleaning and preprocessing
• ├── 2_feature_selection.py # Feature engineering and selection
• ├── 3_model_training.py # Model training and validation
• ├── 4_model_saving_prediction.py # Save models and generate predictions
• ├── 5_customer_segmentation.py # Customer segmentation logic
• ├── 6_automation_script.py # End-to-end pipeline automation
• └── requirements.txt # Project dependencies

Getting Started

Prerequisites

You'll need Python 3.7+ and the packages listed in requirements.txt. The usual suspects are there:

• pandas, numpy for data manipulation
• scikit-learn for machine learning
• xgboost for gradient boosting
• matplotlib, seaborn for visualization

Installation

1. Clone this repository:

• git clone https://github.com/sdlk4/Customer-Lifetime-Value-Prediction.git
• cd Customer-Lifetime-Value-Prediction

2. Install the required packages:

• pip install -r requirements.txt

3. Run the automated pipeline:

• python 6_automation_script.py That's it! The script will handle everything from data processing to final predictions.

How It Works

1. Data Processing (1_data_processing.py)

• Loads and cleans raw customer data
• Handles missing values and outliers
• Creates derived features like recency, frequency, and monetary values

2. Feature Selection (2_feature_selection.py)

• Identifies the most predictive features for CLV
• Removes redundant or low-impact variables
• Prepares the final feature set for modeling

3. Model Training (3_model_training.py)

• Trains multiple regression models (Linear, Random Forest, XGBoost)
• Performs cross-validation and hyperparameter tuning
• Selects the best-performing model based on evaluation metrics

4. Predictions & Model Saving (4_model_saving_prediction.py)

• Saves the trained model for future use
• Generates CLV predictions for all customers
• Creates prediction confidence intervals

5. Customer Segmentation (5_customer_segmentation.py)

• Segments customers into High, Medium, and Low value groups
• Generates segment profiles and characteristics
• Creates actionable insights for each segment

6. Automation (6_automation_script.py)

• Runs the entire pipeline with a single command
• Handles error checking and logging
• Perfect for scheduled runs or production deployment

Key Features

• Automated Feature Engineering: Automatically creates RFM (Recency, Frequency, Monetary) features and other relevant predictors
• Multiple Model Comparison: Tests various algorithms to find the best performer for your data
• Customer Segmentation: Goes beyond predictions to provide actionable customer segments
• Production Ready: Includes model saving/loading and batch prediction capabilities
• Comprehensive Evaluation: Multiple metrics and visualizations to assess model performance

Sample Output

The project generates several key outputs:

• clv_predictions.csv: Customer-level CLV predictions
• customer_segments.csv: Customer segmentation results
• model_performance_comparison.csv: Model evaluation metrics
• Various visualization plots showing model performance and segment characteristics

Model Performance

The models are evaluated using:

• Mean Absolute Error (MAE)
• Root Mean Square Error (RMSE)
• R² Score
• Cross-validation scores Typical performance on retail datasets shows R² scores above 0.75, but your mileage may vary depending on data quality and business context.

Contributing

Found a bug or have an idea for improvement? Feel free to:

• Fork the repository
• Create a feature branch
• Make your changes
• Submit a pull request

Technical Notes

• The project uses XGBoost as the default algorithm due to its strong performance on tabular data

• Feature scaling is handled automatically within the pipeline

• The segmentation uses quantile-based thresholds but can be customized for business needs

• Predicted CLV: $1,245.67