This study presents the constructing procedure, learning process, and application of Artificial Neural-Fuzzy Inference Model (ANFIM). The model combines the traditional neural networks and the fuzzy inference theorem, consequently, it gains more advantages than each one of them. Moreover, our developed model uses the hybrid algorithms of least square estimation and the gradient descent method, it can, hence, adjust the model's parameters effectively. The results of model simulation on a deterministic nonlinear function and the ten-day streamflow of the Dai-Chia Creek basin show that the applicability and practicability of the proposed model.