Discovering Conceptual Differences among Different People via Diverse Structures

• Authors: Tetsuya Yoshida;Teruyuki Kondo;Shogo Nishida
• Venue: PAKDD '99 Proceedings of the Third Pacific-Asia Conference on Methodologies for Knowledge Discovery and Data Mining
• Year: 1999

We extend a method for discovering conceptual differences among people by introducing diverse structures utilizing Genetic Algorithm (GA). In general different people seem to have different ways of conception and thus can have different concepts even on the same thing. Removing conceptual differences seems especially important when people with different backgrounds and knowledge carry out collaborative works as a group; otherwise they cannot communicate ideas and establish mutual understanding even on the same thing. In our approach knowledge from users is structured into decision trees so that differences in concepts can be discovered as the differences in the structure of trees. In our previous approach ID3algorit hm is utilized to construct a single decision tree based on the information theory. However, it has a problem that conceptual differences which are not represented in the tree due to the low information gain cannot be dealt with. To solve this problem, this paper proposes a new method for discovering conceptual differences which utilizes diverse structures via GA. Experiments were carried out on motor diagnosis cases with artificially encoded conceptual differences and the result shows the superiority of introducing diverse structures with GA to a single decision tree which is constructed with ID3.