Energy efficiency modeling and predicting using advanced machine learning

Abstract

In order to push for further energy conservation and greenhouse emission reduction, a hybrid clustering-based prediction approach is proposed to estimate building energy performance. Our proposed method will be examined through the use of a case study, which involves a dataset containing Chicago’s building energy performance. The reported data is collected by the government, with the aim to tracking the cardon dioxide consumption and building energy efficiency. The dataset is first pre-processed through data cleansing and simplification. By combining the density-based spatial clustering of applications with noise (DBSCAN) method with the random forest (RF) method, regression analysis is used to predict the consumption and efficiency in different clusters. This research aims to combine unsupervised and supervised learning methods to predict building energy consumption with increased accuracy.