Comparative study of machine learning approaches for predicting short-term photovoltaic power output based on weather type classification