Optimizing energy efficiency through explainable AI: A genetic algorithm-optimized ensemble method for improved decision-making

Abstract

Accurate prediction of building energy consumption is crucial for optimizing energy usage and costs. This paper proposes an ensemble learning framework optimized with genetic algorithms and explainable AI to predict and interpret energy consumption. The UC Irvine Energy Efficiency dataset containing 768 samples with 8 features is leveraged. The ensemble model combines Random Forest, Decision Tree, and XGBoost regressors, tuned using genetic algorithms. Results show the ensemble model achieves the lowest RMSE of 0.434 and the highest R-squared of 0.998, outperforming individual models. Explainable AI via LIME is also implemented to locally interpret predictions and feature importance. The approach provides improved accuracy and model transparency for energy consumption forecasting. This demonstrates the efficacy of combining diverse machine learning models with optimization and explainability for multivariate time series regression tasks. The tunable ensemble framework with explainable predictions can enable smarter energy planning and decisions.