Utilizing SHAP Analysis to Enhance Machine Learning Models for Predicting Negative Outcomes in Breast Cancer Surgical Procedures

1

A machine learning model using perioperative data effectively predicts adverse postoperative outcomes in breast cancer surgery.
2

The XGBoost algorithm outperformed other models, achieving an AUC of 0.840 in internal validation and 0.780 in external validation.
3

Key predictive factors identified through SHAP analysis included the systemic immune-inflammation index, prognostic nutritional index, and age.
4

The XGBoost model demonstrated good calibration and provided the highest clinical net benefit across various threshold ranges.
5

This study highlights the potential of machine learning and SHAP analysis to enhance clinical decision-making in breast cancer treatment.

1