Improving the accuracy of rainfall-runoff simulations is an important challenge for efficient water resource management. Data-driven models are alternatives for simulating and predicting streamflows based on the relationships between meteorological variables and runoff. To improve runoff forecasting performance, we present data-driven model-based runoff forecasting algorithms coupled with a baseflow separation process. For the evaluation, we used two types of data-driven algorithms, deep neural network (DNN) and random forest (RF), and considered the historical patterns of precipitation, air temperature, humidity, and dam inflows as input data for the algorithms. In addition, we evaluated the prediction model by applying lead times of 1–7 days to construct the optimal input datasets. The performance of the dam inflow prediction using data-driven models coupled with the baseflow separation process was better than that of the algorithm without the process.