Case Mining from Raw Data for Case Library Construction

Authors

Chien-Chang Hsu ⁰, Ye-Hong Huang

Corresponding Author

Chien-Chang Hsu

DOI

https://doi.org/10.2991/jcis.2006.50How to use a DOI?

Keywords

Case library, Feature extraction, Case mining, Genetic algorithm

Abstract

Case-based reasoning systems usually use prior experiences and examples to solve problems. The successfulness of the systems depends on the completeness of case library. It may generate contradictory solutions or increase adaptation cost if the case library contains irrelevant and disordered cases. This work proposes a case mining system to extract representative cases from raw data. The system constructs a case library by feature mining and case mining. Feature mining evaluates relevance between feature and class by fuzzy measurement. The system then uses relevant features to divide raw data into different clusters. Case mining selects cases from each cluster by genetic algorithm. Finally, the system verifies completeness of case library by covering test and utilization statistics. The experimental results show that the system can select representative cases from the data correctly.

Open Access

This is an open access article distributed under the CC BY-NC license.