This paper proposes a practical approach of a neuro-genetic algorithm to enhance its capability of predicting water levels of rivers. Its practicality has three attributes: (1) to easily develop a model with a neuro-genetic algorithm; (2) to verify the model at various predicting points with different conditions; and (3) to provide information for making urgent decisions on the operation of river infrastructure. The authors build an artificial neural network model coupled with the genetic algorithm (often called a hybrid neuro-genetic algorithm), and then apply the model to predict water levels at 15 points of four major rivers in Korea. This case study demonstrates that the approach can be highly compatible with the real river situations, such as hydrological disturbances and water infrastructure under emergencies. Therefore, proper adoption of this approach into a river management system certainly improves the adaptive capacity of the system.