Machine Learning Algorithms for Predicting Electricity Consumption of Buildings

Given that the population is increasing and also energy resources are decreasing, in this study we examine the amount of domestic energy consumption. The purpose of this study is to predict the factors affecting energy consumption in buildings. For this prediction, algorithms of decision tree, random forests and K-nearest neighbors have been used. These algorithms are available in Orange software. In this study, univariate regression algorithm is used to select the best factors. This algorithm identifies the most important factors affecting energy consumption and their impact. The results of this study show that the overall height, roof area, surface and relative compaction have the greatest impact on energy consumption of buildings. The percentage of forecast error for cooling load and heating load are 1.128 and 0.404, respectively. Also, among the tested algorithms, random forest gets the best result.