In this study, stochastic models were developed to forecast on monthly flows of Dim, Manavgat and Koprucay streams. To develop the models 36, 42 and 57 year data were used for 9-13, 9-18 and 9-02 stations, respectively. As the result of the modeling study Auto Regressive model, AR(2) for Dim stream, AR(3) for Manavgat and Koprucay streams were selected. To determine the optimal sthocastic model, Akaike Information Criteria Test was applied. It was also investigated whether residuals were internally dependent by using Portmanteau Test. Synthetic series, at the same length of each historical series were generated by using the selected models. It was observed that synthetic series were well adjusted with historical series. Also, Artificial Neural Network models (ANN) were formed for flows of selected stations. As, results of two methods were compeared, it was determined that AR models were given acceptable results than ANN.