Streamflow prediction using a hybrid methodology based on variational mode decomposition (VMD) and machine learning approaches

Abstract The optimal management of water resources depends on accurate and reliable streamflow prediction. Therefore, researchers have become interested in the development hybrid approaches recent years to enhance performance modeling techniques for predicting hydrological variables. In this study, models based variational mode decomposition (VMD) machine learning such as random forest (RF) K -star algorithm (KS) were developed improve accuracy forecasting. monthly data obtained between 1956 2017 at Iranian Bibijan Abad station Zohreh River used purpose. initially decomposed into intrinsic modes functions (IMFs) using VMD approach up level eight develop models. following step IMFs by RF KS methods. ensemble forecasting result is then accomplished adding IMFs’ outputs. Other models, EDM-RF, EMD-KS, CEEMD-RF, CEEMD-KS, also research order assess VMD-RF VMD-KS findings demonstrated that preprocessing enhanced standalone models’ performance, those performed best terms increasing predictions. model proposed a superior method root mean square error (RMSE = 13.79), absolute (MAE 8.35), Kling–Gupta (KGE 0.89) indices.