ABSTRACT
ABSTRACT: Case retrieval is an important link in Case-Based Reasoning (CBR). In this paper, we choose ontology as a tool to describe cases and construct case library. On this basis, we analyze the features and retrieval requirements to choose appropriate similarity measurement and design the case retrieval algorithm. We proposed a new similarity measurement called Matched Genealogy Measure (MGM), considering not only elements separately but also cases as a whole. Then, we took combat simulation as an example to illustrate the ontology-based algorithm and designed experiments to evaluate it by comparing with some existing measurements. Furthermore, we performed user studies to obtain the ideal retrieval results and compared the retrieval results of different algorithms using Window Distance with the ideal results. The results suggest that MGM could perform better in matching human intuition and obtain reasonable retrieval results. Finally, we adjusted the two coefficients in MGM and demonstrated that MGM could meet different preferences of users.
