Trustworthiness evaluation and retrieval-based revision method for case-based reasoning classifiers

To achieve better classification performance using case-based reasoning classifiers, we propose a retrieval-based revision method with trustworthiness evaluation for problem solving. An improved case evaluation method is employed to evaluate the trustworthiness of the suggested solution after the reuse step, which will divide the target cases and its suggested solutions into a trustworthy set and an untrustworthy set in accordance with a threshold value of trustworthiness. The attribute weights are adjusted by running a genetic algorithm and are used in the second round of retrieval of the untrustworthy set to obtain the classification results. Experimental results demonstrate that our proposed method performs favorably compared with other methods. Also, the proposed method has less computation complexity for the trustworthiness evaluation, and enhances understanding on thinking and inference for case-based reasoning classifiers.