Streamflow forecasting is crucial for managing water resources. A new method was developed to improve daily streamflow predictions by recognizing different flow patterns. By using a combination of techniques like self-organizing maps and deep belief networks, the model was able to accurately forecast streamflow at a hydrological station in China. The integrated framework outperformed traditional models, with higher accuracy in predicting streamflow and peak floods. This approach considers the complexity of different flow patterns, leading to more precise daily streamflow forecasts.