Abstract: In the Cases-Based Reasoning（CBR） system, all new cases are uncritically added to the case base, resulting in case base filled with many redundant and inaccurate cases. To solve this issue, this paper proposes a memory algorithm of case base based on utility value. This algorithm can selectively add new cases to case base, delete redundant and inaccurate cases, strengthen the cases which often judge correctly, and weaken the cases which often judge uncorrectly. The experiments show that, the MCBR system not only can reduce the storage space of case base, but also improve the classification accuracy.

Key words: Case-Based Reasoning（CBR）, utility value, memory algorithm

摘要： 基于范例推理的系统中，如果所有新范例不加辨别地加入到范例库，那么范例库中将充满许多冗余范例和噪声范例。针对这个问题，提出基于效用值的范例库记忆算法。该算法能选择性地加入新范例，删除范例库中冗余的和错误的范例，强化经常判断正确的范例，弱化经常判断出错的范例。实验结果表明，包含记忆算法的范例推理系统不但能减少范例库存储空间，还能提高分类准确度。

关键词: 范例推理, 效用值, 记忆算法