Summary: | 碩士 === 輔仁大學 === 資訊工程學系 === 93 === Successful case-based reasoning systems heavily depend on the completeness of the case library. It is difficult to extract the cases and find the solution from the haphazard case library. It may increase the adaptation cost from the sparse case library with few cases. This paper proposes an intelligent case library construction system for case-based reasoning. The system contains two modules, that is, data analyzer and case extractor. The data analyzer fills missing feature value and deletes irrelevant features according to the frequency, dependency, and importance of features. The data analyzer then project the feature value into two-dimensional feature space to conduct data reduction and data evaluations. The case extractor groups the data of feature space into different clusters according to the coordination in the geometrical space and data covariance. Finally, the case extractor takes the center of each cluster as representative case based on the distance, frequency, and value difference evaluation. The system provides a systematic approach to do data reduction as well as complete case-library construction process. It not only extracts the representative case from the database but also improve the efficiency of case-based reasoning by setting the initial weights of each feature.
|