Comparative Energy Consumption Forecasting Using XGBoost, LightGBM,LSTM ,and ARIMAX with IoT-Based Data Acquisition

Abstract This study presents a comparative analysis of various artificial intelligence techniques to develop an efficient Building Energy Management System (BEMS) capable of predicting energy consumption and contributing to the reduction of energy waste. The research focuses on identifying the most appropriate method for building a predictive model of energy consumption, using both machine learning and statistical approaches. Several methods are evaluated to determine the optimal model for this task, comparing their performance to highlight the strengths and limitations of each one. This work specifically focuses on the comparative analysis of XGBoost, LightGBM,LSTM, and ARIMAX, complemented with an IoT-based data acquisition system designed to capture energy consumption patterns in Moroccan homes. The results demonstrate the superiority of machine learning models, particularly LightGBM, over traditional statistical approaches in terms of accuracy, execution speed, and ability to handle the inherent complexity of energy consumption data