p15 / Conference
A federated explainable ai model for breast cancer classification
European Interdisciplinary Cybersecurity Conference (EICC 2024), 194-201 · 2024
In plain language
This study combines federated learning with explanations for breast-cancer classification on two Wisconsin benchmark datasets.
Research context
It demonstrates a way to study predictive performance and interpretable decisions without centralizing the underlying patient records.
Approach
Artificial neural networks and XGBoost are trained in federated settings. Locally computed SHAP values explain feature contributions.
Findings
Reported accuracy reaches 97.59% on the diagnostic dataset with a neural network and 97.14% on the breast-cancer dataset with XGBoost.
Summary based on abstract.