Retail Sales Forecasting Engine

A reproducible harness for comparing classical, supervised, and foundation-model forecasts across retail demand regimes.

View the code on GitHub

Evidence status

This project intentionally does not present an untraceable reference leaderboard. A valid result consists of the generated metrics and run_metadata.json, which identifies the effective configuration, commit, model labels, and runtime. The default dataset is synthetic; M5 is opt-in.

Try the dependency-light path

python -m pip install -e ".[dev]"
python -m scripts.run_benchmark --classical-only --max-series 50
python -m scripts.generate_report --results results/metrics.csv

Documentation

• Methodology
• Models
• Metrics
• Reproducibility
• Deployment

The repository is an experimental and educational tool, not a production demand-planning system or an official M5 leaderboard implementation.