Forecasting electricity demand with ML
A hands-on experiment: how well can simple machine-learning models predict GB electricity demand from weather and calendar features?
Electricity demand is famously predictable — until it isn’t. It tracks the weather, the calendar and human routine in ways that look like an ideal playground for machine learning.
The idea: take public GB demand data, build up from a naive baseline through to gradient- boosted and sequence models, and show — interactively — where each one wins, where it fails, and what features actually matter. As much about understanding the system as chasing accuracy.
On the roadmap.