Assessment by belief
Takao Miura, Isamu Shioya · Australasian Database Conference · 2001
We discuss how to post-evaluate inductive classification based on user belief. Although we could learn classification rules inductively by means of decision tree generation, we wonder whether it is consistent with our utilization or not. In the investigation we discuss how to obtain assessment of learning results by verifying belief. Our idea is based on a decision tree with hierarchy to class and attributes; to each attribute we assume taxonomy on the domain in addition to class hierarchy. Then, given a firm belief (such as regulation and top executive policy), we check whether the trees satisfy it and we can see the usefulness of the trees.