Extracting Diagnostic Knowledge from Hepatitis Data by Decision Tree Graph-Based Induction

摘要

Decision Tree Graph-Based Induction (DT-GBI) is a technique for constructing a decision tree from graph-structured data. In DT-GBI, substructures (discriminative patterns) are extracted by stepwise pair expansion (pair-wise chunking) and used as test attributes at nodes of a decision tree. We applied DT-GBI to a classification task of hepatitis data. In the first experiment, the stages of fibrosis are used as classes and a decision tree is constructed for discriminating patients with F4 (cirrhosis) from patients with the other stages using only the tune sequence data of blood inspection. In the second experiment, the types of hepatitis (B and C) are used as classes and a decision tree is constructed by DT-GBI as in the first experiment. The preliminary results of experiments, both constructed decision trees and their predictive accuracies, are reported in this paper.