Evolutionary Conformal Prediction for Breast Cancer Diagnosis

Conformal Prediction provides a framework for extending traditional machine learning algorithms, in order to complement predictions with reliable measures of confidence. The provision of such measures is significant for medical diagnostic systems, as more informed diagnoses can be made by medical experts. In this paper, we introduce a conformal predictor based on genetic algorithms, and we apply our method on the Wisconsin breast cancer diagnosis (WBCD) problem. We give results in which we show that our method is efficient, in terms of accuracy, and can provide useful confidence measures.