Seasonal prediction of chemical components in trans-regional silver carp (Hypophthalmichthys molitrix) based on random forest and SHAP interpretation models

Variations in the chemical composition of silver carp are recognized as one of the major factors responsible for quality fluctuations during surimi processing. This study systematically investigated the effects of region, month, and size on the basic chemical components of silver carp ( Hypophthalmichthys molitrix ). A chemical component prediction model was constructed using the Random Forest algorithm, and the SHapley Additive exPlanations (SHAP) model was integrated to interpret multi-factor interaction mechanisms. Results showed: (1) Silver carp chemical components exhibited significant region-season-size dependence: small-sized silver carp from Taizhou (Jiangsu) had the highest crude protein (90.52 g/100g in Mar, dry-weight basis) and salt-soluble protein (62.72 g/100g in Dec., dry-weight basis) among the three regions, while large-sized individuals from Anshan (Liaoning) had significantly higher winter fat content (34.47 g/100g in Jan., dry-weight basis) than other regions; (2) Across 10 repeated stratified random partitions, the RF model achieved R² values of 0.884±0.023, 0.917±0.023, and 0.866±0.027 for predicting moisture, fat, and salt-soluble protein contents, respectively; (3) SHAP analysis indicated region-, month-, and body-weight-dependent contribution patterns, including positive model contributions to salt-soluble protein prediction in Qianjiang from January to June and body-weight-related moisture variation in Taizhou.