Fuzzy-Enhanced Smart Building Energy Forecasting: A Case Study from the 2025 GECAD Competition

This paper presents a fuzzy-based approach for electric energy consumption forecasting in the context of the 2025 GECAD Competition on Electric Energy Consumption Forecast Adopting Multi-criteria Performance Metrics. The proposed methodology leverages fuzzified temperature features using triangular membership functions combined with time-ofday data to train and evaluate multiple regression models, including Random Forest, XGBoost, Support Vector Regression (SVR), Multilayer Perceptron (MLP) regressor, and Linear Regression. Experimental results demonstrate that the XGBoost model achieved the best root mean square error (RMSE) performance, with a value of 1601.59 on the test dataset, while the MLP regressor obtained the best mean absolute error (MAE), with a value of 1097.01. These findings highlight the effectiveness of fuzzy-enhanced feature engineering in improving the accuracy of energy consumption prediction tasks across different performance metrics.