ltv-predictor

Predict customer lifetime value by computing RFM (Recency, Frequency, Monetary) features from raw order data, then training regression models—linear or random forest—to estimate LTV. This Skill automates the entire pipeline: load CSV order files, engineer RFM features, train and compare models with cross-validation, and generate batch predictions for customer segments.

RFM analysis segments customers by Recency (last purchase), Frequency (purchase count), and Monetary value (total spent). These three dimensions become features that predict future customer lifetime value, enabling data-driven lifecycle management and targeted retention strategies based on purchase behavior patterns.

Yes. After training models and predicting LTV, this Skill exports automated reports in HTML, Markdown, and Excel formats. Reports include predictions, model comparisons, visualizations, and business metrics, reducing manual reporting time for retail and e-commerce contexts.

The LTV predictor accepts CSV order data and handles preprocessing internally: data cleaning, outlier handling, and feature scaling. It requires order-level records with transaction details to compute RFM features; openpyxl and pandas manage input validation and output formatting.

This Skill trains both linear regression and random forest models on RFM features and compares performance using cross-validation. Random forest typically captures non-linear relationships in customer behavior; linear regression is simpler and more interpretable. Choose based on accuracy, interpretability, and business needs.

You need Python with pandas, numpy, scikit-learn, matplotlib, seaborn, openpyxl, and flask installed. The Skill runs as scripts in a Python environment; these dependencies handle data manipulation, model training, visualization, and reporting—no additional setup beyond a standard data science environment.