Feature Selection and Optimization for Water Quality Index Design in the Rachiine River

This study develops an improved Water Quality Index (WQI) for the Rachiine River by integrating objective feature selection with the CCME-WQI framework. From an initial set of 20 physicochemical, microbial, and heavy metal parameters, a reduced subset is identified using correlation analysis, and confirmed through PCA and EFA. Feature importance is further assessed using Fisher score, SHAP values, and discriminant analysis. A genetic algorithm is then applied to optimize parameter subsets that approximate the full WQI while preserving ecological and statistical relevance. The resulting index achieves strong performance with reduced complexity, supporting cost-effective and interpretable water quality monitoring.