🚲 mle-bike

Forecasting bike-share availability one hour ahead, continuously, on live open data — an end-to-end ML engineering showcase. Every prediction meets its ground truth an hour later.

Predictions vs. reality

Bikes available at one station: what the model said an hour in advance, what the persistence baseline said, and what actually happened.

Actual Model forecast Baseline (persistence)

Is the model earning its keep?

Hourly mean absolute error across all stations — the model must stay below the baseline, or the weekly retrain won't promote it.

Model MAE Baseline MAE