A rule based case maintenance method for the performance of CBR classifier

Aiming at the decreasing efficiency and learning ability problem of CBR classifier that the growing scale of case base brings, a case maintenance strategy based on selective rules is proposed to delete the redundant and noisy cases. First, as to remove the noise information that lies in the feature attribute, a feature reduction method based on information entropy and GA is proposed to. Then, three specific rules are defined to select out the cases that are satisfied with the rules, and delete those are not to accomplish the maintenance process. The experimental results indicate that the proposed method achieves a better performance of CBR classifier and decreases the time complexity at the same time, which prove that the proposed could remove the redundant and noisy cases, and guarantee the learning ability of the CBR classifier.