Construction of a VIKOR Composite Index-Based Probabilistic Prediction Model for Urban Water Blackening and Odor Recurrence Using Multi-Method Feature Selection

• • Established a power-law relationship between transparency and turbidity (y = 3.12x−0.66), setting a critical turbidity threshold of 46.9 NTU for blackening and odor, enabling turbidity as a surrogate for transparency in monitoring. • • Identified five core indicators (turbidity, DO, TP, CODMn, NH3-N) via ANOVA, RFE, and RF, with turbidity as the most important, providing a parsimonious yet comprehensive set for predicting blackening and odor. • • The VIKOR-based composite index model achieved high predictive accuracy (RMSE = 0.029, MAE = 0.020) and consistency (NSE = 0.918, R2 = 0.918), outperforming other indices (R2 < 0.84), ensuring reliable early warning. • • The model exhibited robust performance across multiple river basins (Yangtze, Pearl, Haihe, Yellow), indicating its generalizability for nationwide application in urban water management.