摘要： 案例推理方法建立在“相似问题具有相似解”的基础上，能否从案例库中检索出与新问题“最相似”的案例是案例推理方法成功的关键因素之一。该文提出一种改进的检索方法，在原始最近相邻算法基础上，用专家对新问题案例与历史案例属性差异的效用评价替代原始的属性差异值来衡量专家对属性差异的敏感程度。引入变异系数来标度新问题案例与历史案例的属性差异的分布情况，从而保证检索出的最相似案例具有较高的属性差异的均衡性。通过具体案例检索实例分析，验证了该方法的有效性。

关键词: 案例检索, 偏好信息, 案例推理

Abstract: Case-Based Reasoning(CBR) is a problem solving technique that solves new problems by finding a similar past case, and reusing it in the new problem situation. Noticing the shortcomings of K-Nearest Neighbors algorithm, this paper proposes an improved case retrieval algorithm measuring the similarity of new problem case and historical cases of case-based not only by difference of their feature values, but also by attribute(sensitivity) of reasoner. In this improved algorithm, original distance of feature of two cases is replaced by the reasoner’s utility value of the distance value. In addition, a variance indicator is introduced to balance equilibrium degree of all differences. The proposed algorithm is compared with other popular algorithms and verified numerically.

Key words: case retrieval, preference information, Case-Based Reasoning(CBR)

中图分类号: 

• TP182