Predicting Energy Consumption in Educational Buildings: A Comparative Study of Machine Learning Models

The increasing energy consumption in educational institutions underscores the need for accurate predictive models to enhance energy management and sustainability. This study compares the effectiveness of multiple linear regression (MLR), artificial neural networks (ANN), deep neural networks (DNN), and support vector regression (SVR) in predicting energy consumption in educational buildings. The dataset included variables such as humidity, temperature, occupancy, and power consumption. Each model’s performance was evaluated using key metrics, and optimal configurations were identified. The results show that DNN models, with their superior ability to capture complex patterns, achieved the highest predictive accuracy, followed by SVR, ANN, and MLR. The study also highlights the significant impact of the number of neurons in ANN and DNN models, as well as the regularization parameter and kernel parameter in SVR models. These findings demonstrate the potential of advanced machine learning techniques to improve energy consumption predictions and contribute to more efficient energy management in educational buildings. Future research should focus on further optimizing these models and exploring hybrid approaches for enhanced predictive accuracy.