Comparative study of multi-combination models for medium - and long-term runoff prediction in Weihe River

The characteristics of hydrological data include nonconsistency and nonlinearity. prediction accuracy can be improved through the combination both decomposition algorithm runoff model. Previous studies have typically focused on a single These compared before after decomposition, ignoring role multiple algorithms models. Considering limitations previous combinations models, this study will explore unique features by using five algorithms, including Empirical Mode Decomposition (EMD), Ensemble (EEMD), TIME series (TIME), Variational (VMD), Singular Spectrum Analysis (SSA). constructed models for Prophet, Long Short-Term Memory (LSTM), Multiple Regression (MLR), Random Forest (RFR), Gradient Boosting (GBR), Support Vector (SVR). Thirty combined were then developed used to forecast medium long-term at Xianyang Station. To comprehensively evaluate forecasted results, evaluation metrics used. EMD but difference was insignificant, least effective. VMD, EEMD, SSA, other hand, yielded higher quality. model achieved an NSE above 0.70, demonstrating good results. Of thirty SSA-SVR SSA-LSTM most accurate, with verification 0.90. This comprehensive, reliable, accurate employing findings provide framework characteristics-driven watershed water resources scheduling.