Bridging the Gap Between the Physical-Conceptual Approach and Machine Learning for Modeling Hydrological Systems