Climate change has significantly increased the frequency and intensity of heat waves in the world, especially in the western regions of Iran. In fact, heat waves are one of the most important climate hazards that leave numerous harmful consequences in human life and other living organisms every year. This phenomenon has significant impacts on public health, agriculture, and energy consumption, and highlights the need to develop accurate forecasting and early warning systems. This research was conducted using daily data on temperature, relative humidity, and climate indicators at meteorological stations in the five western provinces of Iran (Kermanshah, Hamadan, Ilam, Kurdistan, and Lorestan) during the period 2013 to 2023 and additional ERA5 satellite data, Random Forest, LSTM, and simple neural network models. The performance of the models was evaluated with RMSE, MAE, and Accuracy indices. The results showed that the Random Forest model performed best with 92.7% accuracy and the lowest error (RMSE=1.42). Also, using the SHAP method, the main factors of heat wave formation were identified, with daily maximum temperature (37.2%), relative humidity (21.4%), and solar radiation (16.8%) having the greatest impact. Finally, a framework for an early warning system was designed that can provide local warnings 5 to 7 days before the occurrence of heat waves and play an effective role in the optimal management of water, energy, and public health resources.