Project case study
Revenue Compass
Validate daily revenue data and compare predictive models against a simple baseline.
The approach
Validates CSV input and compares XGBoost, CatBoost, LightGBM and Extra Trees across multiple backtest windows. A model is published only when it beats the seven-day baseline, then produces 7–90 day forecasts.
Results and scope
What the project demonstrates
An example Rossmann evaluation reports 8.92% WAPE over 135 backtest days and 70.08% MAE improvement over the baseline. These results are dataset-specific.