A Proposed Method to Handle Classification Uncertainty Using Decision Trees

Abstract

A novel method is proposed in this paper to handle the classification uncertainty using decision tree classifiers. The algorithm presented here extends the decision tree framework to give the ability of measuring the confidence of the classification. Using this algorithm a certain number of the input samples are rejected as "risky points" in order to obtain a smaller misclassification rate on the remaining points. The algorithm is being integrated into a Medical Decision Support System where a confmdence number to every classification is required.