Mcompetitions/M4-methods ? reverse-engineered prompt
Reverse engineered prompt
Build me a small project for the M4 forecasting competition that lets me explore the competition data, reproduce the benchmark forecasts, and compare different methods in one place.
I want the dataset loaded with the train and test series plus the series info like frequency, horizon, seasonality, and start date. I also want the common evaluation metrics for forecast quality, especially sMAPE, MASE, and OWA, so I can score point forecasts and rank methods by frequency and domain.
Please include the benchmark forecasting code, the 4Theta method, and support for the saved point forecasts and prediction intervals so I can inspect results from the submitted methods. If any of the code needs current package guidance, look up the latest docs online first.
Make it easy to run the main scripts from a clean setup and to see the final comparison tables or ranks without having to dig through the folders.
Are you gonna build this?
make sure you review the code using coderabbit