Electric Power Consumption Prediction from Scarce Dataset with Entropy-Based Subset Selection Regression Tree (e-ss RT)

In this paper, the problem of predicting the electrical power consumption and deriving a monthly power consumption plan for a building from a scarce dataset is considered. To solve this problem, a novel prediction method is introduced based on the concept of entropy-based subset selection and the regression tree algorithm to generate a precise model for predicting the power consumption, provided that there are numerous missing data points in the dataset. An experimental setup has shown that the proposed algorithm can precisely predict the power consumption inside a real structure with an accuracy up to 92%. Furthermore, the algorithm can provide information whenever there is an indication of power waste inside the structure.