SHAP Explainability
Glucose, BMI, and Age were the dominant predictors in Random Forest SHAP attribution. Symptoms and some additional variables contributed little after glucose and BMI were available. High glucose increased predicted diabetes probability. SH…
1 sources - 5 claims
Glucose, BMI, and Age were the dominant predictors in Random Forest SHAP attribution. Symptoms and some additional variables contributed little after glucose and BMI were available. High glucose increased predicted diabetes probability. SHAP was used to explain the strongest tree-ensemble diabetes model selected by cross-validated AUC. BMI had a bimodal contribution in which low values were protective and elevated values increased risk.