pyaf/load_forecasting

Implements nine time series forecasting algorithms—from classical methods (ARIMA, exponential smoothing, moving averages) to deep learning approaches (LSTM, GRU, RNN)—with automated hyperparameter tuning via grid search. Features automated data pipelines that scrape real-time load and weather data from Delhi's grid operator and weather services, then feeds them through a Django web interface for daily forecasts and cross-model performance comparison. Includes production scheduling scripts that run on AWS to generate daily predictions across all models using rolling windows of historical data.

620 stars.