The advancement of smart meters using evolving technologies such as Internet of Thing (IoT) is producing more data for the training of energy prediction model. Since many machine learning techniques were not premeditated for handling large number of irrelevant features, it has engendered the search for optimal techniques to decrease the generated features and potentially identity the most pertinent or relevant features that have impact on building energy efficiency.feature selection is receiving more attention in the machine learning field. However, in the building energy prediction domain, this attention has been focused more on unsupervised learning (clustering) without much attention or examination of feature selection for supervised learning (classification). To address this gap, this study focuses on the application of feature selection for supervised learning using different sizes of data. This study conducts a comparative analysis of 14 machine learning algorithms on 5 different data sizes and explored the effect of 7 feature selection methods on model performance for predicting energy consumption in buildings. Filter method was noted as the most suitable method for most ML classification algorithms. Also, the most effective model for predicting energy performance in buildings was identified as Gradient Boosting (GB).