Integrating Multi-Layer Perceptron Regression with Innovative Optimization for Accurate Building Cooling Load Prediction

This study explores the application of a machine learning model called Multi-Layer Perceptron Regression (MLPR) for building cooling demand prediction. Through a hybridization technique with two cutting-edge optimization algorithms, the Brown Bear Optimization Algorithm (BBOA) and the Non-Monopolize Search Algorithm (NMSA), it sets out to explore its optimization potential. This leads to the development of optimized MLTN and MLBB models. The dataset was divided into 70% for training and 15% each for the rounds of testing and validation to make sure that the assessment is robust. Five insightful evaluation measures were utilized to assess the performance of the models, namely: MARE, MSE, RMSE, R2, and NRMSE. A model with the highest R2 and lowest error metric values across all phases is considered superior. Further, careful analysis of the layers of all three models constantly shows that the MLBB model is better compared to others, as seen by the highest R2 and lowest error values it has. In the third layer, the MLBB model gave a good performance, with an R2=0.999, RMSE=0.360, MSE=0.130, MARE=0.015, and NRMSE=0.001, which is really commendable. This goes to heighten the reliable effectiveness of MLBB in making precise predictions of building cooling load, hence it can be applicable in real life for encouraging energy-efficient building operations.