Explainable machine learning model for identifying key risk factors in congenital heart disease prediction using questionnaire data: a retrospective case-control study.

Notably, some features such as maternal BMI and emotional stability, while ranked highly by SHAP, did not reach statistical significance in the regression analysis.

also in 111,027 other papers

This discrepancy reflects the differences between machine learning and traditional statistical approaches: SHAP values quantify a feature’s marginal contribution to model predictions, often capturing non-linear effects and interactions that may not be significant in linear models.

also in 2,774 other papers