Abstract: To resolve deficiencies of existent methods as requirement of larger storage and no consideration on the situation when new cases appear, an improved method to determine weighting coefficient for case feature based on attribute importance theory was proposed. By introducing fission matrix, the relationship between new core, reduction of attribute set and old ones was summarized, and incremental updating algorithms of core and attribute reduction were given. Computational algorithm description of weighting coefficient was illustrated according to above conclusions. Performance was showed by comparison of time and space complexity.

Key words: case-based reasoning, weight, rough set, attribute reduction, attribute importance

摘要： 针对已有案例特征权重确定方法所需存储空间大且未解决新增案例情况下的权重更新问题，提出一种基于属性重要度的改进方法，通过引入分体差别矩阵，发现新的属性核、约简与原有结果之间的关系，设计属性核、约简的增量式更新算法，并给出计算特征项权重的算法描述，利用时间及空间复杂性对比，体现了算法的性能优势。

关键词: 案例推理, 权重, 粗糙集, 属性约简, 属性重要度