Energy Consumption Forecasting in Smart Homes Using LSTM and XGBOOST Ensemble

Abstract: Accurate prediction of residential electricity demand is essential for energy conservation, cost optimization, and effective grid planning. Smart homes generate large volumes of fine-grained consumption data, making them suitable candidates for advanced predictive modeling . This study proposes a hybrid forecasting framework that integrates Long Short-Term Memory (LSTM) networks with Extreme Gradient Boosting (XG Boost). LSTM captures temporal dependencies in consumption sequences, while XG Boost model nonlinear relationships in engineered features. The ensemble produces stable and adaptive predictions suitable for dynamic household environments. A web-based interface supports data upload, real-time forecasting, visualization, and cost estimation. Experimental results demonstrate that the hybrid model consistently outperforms standalone approaches in RMSE, MAE, and MAPE. The system provides interpretable predictions using feature-attribution techniques, enabling users to understand consumption drivers. This research contributes a practical and extensible solution for smart home energy management.

Keywords: Smart Home Energy Forecasting, LSTM, XG Boost, Hybrid Ensemble Model, Deep Learning, Gradient Boosting, Smart Grid Optimization, Demand Response, Feature Engineering.