Oxygen plays a vital role in maintaining the balance of life cycles in all ecosystems. Aquatic life is highly sensitive to dissolved oxygen (DO) levels. This necessitates not only continuous monitoring of DO in aquatic environments but also the development of accurate predictive models for future DO concentrations. The aim of this study is to develop a precise prediction model for DO concentration in river water. To this end, a novel hybrid intelligent model based on the support vector regression (SVR) approach was developed to predict dissolved oxygen levels. Three optimization algorithms—Firefly, Gray Wolf, and Bat—were employed to enhance the modeling of DO in river water. The study utilized hydrometric data from the Molasani station on the Karun River in Khuzestan Province as a case study, covering five combined input parameter scenarios over the period 2012–2022. Model performance was evaluated using correlation coefficient (R), root mean square error (RMSE), mean absolute error (MAE), and Nash–Sutcliffe efficiency coefficient (NSE). Results demonstrated that combined input scenarios improved model performance across all models. Among them, the SVR-Firefly hybrid model achieved the best validation results with a correlation coefficient of 0.970, RMSE of 0.668 mg/L, MAE of 0.520 mg/L, and NSE of 0.975. Overall, the findings indicate that intelligent models based on the support vector regression approach provide an effective tool for sustainable river engineering management.