Predicting Energy Consumption in Commercial Buildings using its Property features and Machine Learning Algorithms

Population growth is a very crucial factor that leads to an upsurge in demand for residential services and lavishness, which is triggering the depletion of energy assets. A stable surge has been noticed from 20% to 40% worldwide towards energy consumption collectively from both residential and commercial buildings. In the past few years, many regression algorithms have been used for predicting energy consumption in buildings. The state-of-the-art for energy consumption prediction in commercial buildings using its property features was 82% (r2 score) based on the topical studies. This study focuses on building a novel multi-class classifier that will categorize energy consumption prediction in multinomial classes using the Commercial Buildings Energy Consumption Survey (CBECS) dataset. This research compares four classification algorithms, namely Gaussian Naïve Bayes, Random Forest, K-Nearest Neighbour and Logistic Regression using analysis of variance (ANOVA) and principal component analysis(PCA). Accuracy, precision, recall, and f1 score are considered as evaluation metrics for this study. Amongst all the classification algorithms, K-Nearest Neighbor achieved the best efficiency of 97% for both ANOVA and PCA, followed by Random forest and Gaussian Naive Bayes . The accuracy of each model was evaluated using the 10-Fold Cross Validation technique. Keywords: Energy consumption, Commercial Buildings, Property Features, Buildings Energy Consumption Surve