Combining load forecasts from independent experts

The NPower Forecasting Challenge 2015 invited students and professionals worldwide to predict daily energy usage of a group of customers. The BigDEAL team from the Big Data Energy Analytics Laboratory landed a top 3 place in the final leaderboard. This paper presents a refined methodology based on the implementation during the competition. We first build the individual forecasts using several forecast techniques, such as Multiple Linear Regression (MLR), Autoregressive Integrated Moving Average (ARIMA), Artificial Neural Network (ANN) and Random Forests (RF). We then select a subset of the individual forecasts based on their performance on a validation period, a.k.a. post-sample. Finally we obtain the final forecast by averaging the selected individual forecasts. The forecast combination on average yields a better result than the forecast from a single technique.