A Hybrid Tree-Based Ensemble Learning Model for Day-Ahead Peak Load Forecasting