Interpretable machine learning analysis of routine blood biomarkers and derived indicators for predicting coronary heart disease in patients with carotid stenosis.

Although the RCS analysis suggested a significant interaction, the SHAP subgroup distribution test did not reach statistical significance ( P = 0.203), suggesting that the smoothing properties of the machine learning model when dealing with non-monotonic relationships may mask some of the true interaction effects [ 28 ].In addition, the SHAP dependency plot shows that alcohol consumption has a moderating effect on the FPG effect: moderate drinkers have a reduced risk in the moderate glycemic range.

also in 111,027 other papers

Figure 5 E shows that NLR shows an overall trend of higher SHAP values with higher NLR, but the relationship is not strictly linear.

also in 4,871 other papers