STREAMFLOW TIME SERIES FORECASTING WITH HYBRID MACHINE LEARNING MODEL

Streamflow forecasting plays a critical role in hydrological studies. Recent advancements in machine learning offer promising alternatives with lower computational requirements, making them well-suited for real-time forecasting. This study investigates the potential of machine learning models to improve streamflow prediction accuracy in data-limited conditions. Addressing challenges posed by incomplete datasets and a scarcity of diverse observational records, the research examines various model structures and methods, with a focus on hybrid models that integrate both recursive and non-recursive components. Results indicate that such hybrid models outperform conventional machine learning models, providing a viable approach for achieving better predictions in data-scarce environments.