An Explainable Machine Learning Framework for Risk Factor Analysis in Healthcare: A Case Study on Breast Cancer
Year: 2026
Type: conference-paper
Abstract: This study presents a comprehensive and interpretable framework for breast cancer risk factor analysis that combines the predictive capability of machine learning (ML) with the transparency of Explainable Artificial Intelligence (XAI). Using the Wisconsin Diagnostic Breast Cancer dataset, three classification models—Logistic Regression, Random Forest, and XGBoost—were evaluated. The XGBoost model achieved the highest performance, attaining an accuracy of 97.3%, consistent with state-of-the-art benchmarks. To overcome the critical limitation of model opacity, the SHAP (SHapley Additive exPlanations) framework was applied to generate feature-level insights and enhance interpretability. Distinct from previous SHAP-based approaches, the proposed framework integrates multi-model comparison with local interpretability validated against oncological literature. The analysis identified mean radius, perimeter mean, and concave points mean as the most influential predictors of malignancy, aligning with established histopathological and sonographic indicators of tumor morphology and aggressiveness. These findings demonstrate that high-performing, transparent AI systems can serve as reliable decision-support partners for clinicians, improving diagnostic confidence and supporting a human-in-the-loop paradigm for clinical decision-making.
Author: Napat Sukthong
Institution: Mahasarakham University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7167347697