Prediction of river discharge of Kesinga sub-catchment of Mahanadi basin using machine learning approaches

River discharge is a relevant ingredient of the hydrological cycle for a wide scale of utilizations and evaluation of water assets, plan of water-related designs and flood admonitory and relief plans. The predictive discharge of the basin using the machine learning approaches is therefore significant for managing water resources and the prevention of flooding control. This investigation evaluated the viability of several machine learning methods, M5P tree, Random forest, Regression tree, reduced error pruning tree, Gaussian process and support vector machine, to predict the basin discharge of the Kesinga basin. Various statistical measures, i.e. correlation coefficient, mean absolute error, root mean square error, Willmott’s index, Nash–Sutcliffe efficiency coefficient, Legates and McCabe’s index and normalized root mean square, error were utilized to assess the performance of the developed model. The presentation of random forest and M5P models was found to be the best when compared with the regression tree, reduced error pruning tree, Gaussian process and support vector machine–based models. Overall RF-based model gave the best results among all applied models for predicting water discharge for the Kesinga basin with the coefficient of determination (R2) values of 0.978 and 0.890 for the training and testing stages, respectively. The main significance of soft computing techniques is that they help users solve real-world problems by providing approximate results that conventional and analytical models cannot solve.