Mesab / Project portfolioInteractive simulation
Machine learning · ANN & SVR

Energy consumption forecasting

Compare forecasts from an ANN and SVR trained on a generated electricity-demand series.

Interactive simulation · Synthetic data / simulated hardware
Synthetic actual demandANN forecastSVR forecast
Hours 0–167 · training historyHours 168–191 · held-out test

Both models are trained in Python on lagged demand and hour-of-day features. The test set is held out chronologically; each one-hour prediction uses previously observed actual demand. Results shown here belong to this recreation and do not represent the original project.

Forecast comparison

Test hourActual kWhANN kWhSVR kWh
Download project source

Includes the browser simulation, implementation notes and supporting example source.