Supplementary Tables S1–S10 accompanying the manuscript "Machine Learning–Based Analysis of Test-Day DHI Variables Concurrently Associated with Elevated Somatic Cell Count in Holstein Cows: Evidence for SNF–Protein Residual as a Lactose Proxy"

Supplementary Tables S1–S10 accompanying the manuscript "Machine Learning–Based Analysis of Test-Day DHI Variables Concurrently Associated with Elevated Somatic Cell Count in Holstein Cows: Evidence for SNF–Protein Residual as a Lactose Proxy" Table S1: SCC threshold sensitivity (150–400 × 10³ cells/mL)Table S2a–S2b: Variance inflation factors (16-variable and 14-variable models)Table S3: Fat/fp_ratio sensitivity analysisTable S4: Hyperparameter sensitivity (LR, RF, XGBoost)Table S5: Leave-one-farm-out cross-validation (37 farms)Table S6: SHAP importance across five lactation stagesTable S7: SHAP importance with cow-level SCC history includedTable S8: Imputation sensitivity analysisTable S9: Farm-level descriptive statistics (37 farms)Table S10: XGBoost and random forest hyperparameter sensitivity (24 configurations) Data derived from 72,547 test-day records from 3,562 Holstein cows across 37 Korean dairy farms (2021–2025). Raw DHI records are not included; they are the property of the Korean Dairy Cattle Improvement Center (DCIC) and are available upon request from www.dcic.or.kr.