Integration of Genetic Algorithm and Support Vector Machine for Water Discharge Prediction in River Basin System

Water discharge (WD) is essential component of river hydrodynamics. It is crucial to predict WD for constructing reservoirs and dams, operations, management and river engineering studies. WD includes various complex nonlinear processes which depends on many variables. The WD prediction is complicated using traditional methods because these are unable to handle the complex nonlinear processes of it. WD prediction using novel artificial intelligence (AI) techniques is highly encouraged for its effectiveness in flood forecasting. A hybrid AI approaches is useful for WD prediction in Mahanadi River (MR), India. Support vector machine (SVM) is integrated with Genetic algorithm (GA) (GA-SVM) for optimization of all associated SVM parameters. Temperature(T), rainfall (R), water level (WL) and sediment yield (SY) are used as inputs parameters for prediction of WD. The standard statistical measures are used for the performance evaluation of GA-SVM model. Hybrid GA-SVM with automated parameter tuning model is developed for prediction of WD at Tikarapara location. It is outermost downstream station in the MR using hydro-climatical parameters. The developed GA-SVM model is compared to SVM, and multiple linear regression (MLR) to evaluate the ability of the models to predict the WD. The proposed GA-SVM model is provided satisfactory and better results for estimation of WD in the MR among all comparative models. This modelling approaches can be applied for the prediction of WD at gauge or ungauged places in MR due to its easiness in implement and satisfactory results.