Machine Learning Techniques for Energy Efficiency Prediction: A Comparative Studies

This paper looks into the energy efficiency of buildings through predicting the heating and cooling load requirements of buildings. The aim of measuring energy efficiency is to minimize the amount of energy to provide better services and reduce energy cost. The need to have robust and accurate prediction system for energy efficiency cannot be overemphasized. In this work, three machine learning techniques, which are: Support Vector Machine, Decision Stump, and Radial Basis Function Networks were applied. The problem is a regression task with two distinct target variables that include heating and cooling load requirements of a building. The performance of the proposed techniques was measured using correlation coefficient, relative absolute error, and root mean squared error. Adequate comparative studies were carried out among the three proposed techniques. Empirical results from this work indicated that Support Vector Machine achieves best results for both target variables, while Decision Stump performed the least.