Decision Tree Learning Approach for Knowledge Acquisition from A Medical Database.(Dept.E)

Abstract

Automated knowledge acquisition (AKA) tools hold promise of preventing the process of acquiring knowledge from becoming the bottleneck in the development of knowledge base systems (KBSs). Most hospitals keep detailed records of patients containing descriptions of symptoms, diagnosis, prescribed medicine, and observed changes over time. Cumulatively, these records contain a wealth knowledge about medical diagnosis and treatment. The knowledge acquisition process would be significantly simplified if an automated system could look at these records, and extract valuable pieces of knowledge. Doctors would only need to verify the output of these knowledge acquisition tools rather than sit through hours of interview with knowledge engineers. This paper presents a new methodology for knowledge acquisition from databases. The p roposed methodology coincide with RITIO algorithm 13) in the rule induction without drawing the decision tree and in the eliminating less effective attributes. It builds the decision tree when needed. The new methodology differs from the ID3' likes algorithms (15) and RITIO since it Gods the global optimal solutions via back tracking. It is tested on standard example and applied on a real world database.