中文摘要：

对于CBR中的案例检索问题，结合经典案例相似度计算方法，对目前在各实际系统中应用最为广泛的k-NN算法进行改进。经过特征约简，在假设时间因素对历史案例可采纳程度有显著影响基础上，提出了一种小规模的基于时序的案例特征权重多阶段调整算法。该算法适用于数值型特征项相似度计算。

英文摘要：

Following the classical approaches of case similarity calculation in CBR retrieval,this paper improves the traditional algorithm of k-NN.After feature reduction ,based on the hypothesis that time factor has a significant affect on the adoptability of the history cases,a small scale algorithm for case feature weight calculation called TSBMPSA is proposed.The algorithm is suitable for numeric features.