Energy Consumption Prediction Using Data Reduction and Ensemble Learning Techniques

Building energy problems have various kinds of aspects, one which is the difficulty measuring efficiency. With current data development, efficiency measurements can be made by developing predictive models to estimate future building needs. However, with massive amount data, several arise regarding quality and lack scalability in terms computation memory time modeling. In this study, we used reduction ensemble learning techniques overcome these problems. We numerosity reduction, dimension a LightGBM model based on boosting added bagging technique, compared incremental learning. Our experimental results showed that could speed up training process prediction without reducing accuracy. Testing also revealed had best performance RMSE speed, an 262.304 1.67 times faster than