With the increasing energy consumption of buildings, accurate prediction of heating and cooling loads has become essential for improving HVAC system performance and supporting efficient energy management. In this study, the performance of eight machine learning algorithms, including Random Forest, XGBoost, Gradient Boosting, Decision Tree, K-Nearest Neighbors (KNN), Support Vector Regression (SVR), Artificial Neural Network (ANN), and Linear Regression, was comprehensively evaluated using three independent datasets comprising a residential building (768 samples) and two office buildings located in the United States (64,792 samples) and Denmark (55,029 samples). The first dataset was employed to validate the proposed framework against a reference study. In addition, energy and exergy analyses were conducted to assess the building's energy performance. The results demonstrated that Random Forest consistently achieved the best performance, with coefficients of determination (R²) exceeding 0.95 across all datasets. Feature importance analysis identified indoor temperature, outdoor temperature, relative humidity, and solar radiation as the most influential variables affecting thermal loads. The proposed framework provides a practical approach for thermal load prediction, building energy performance assessment, and intelligent energy management.