Abstract: Dimension disaster in the classified function will directly impact on the efficiency and the rate of accuracy of K-Nearest Neighbor algorithm（KNN），this paper combines information entropy theory used to reduce attributes and KNN algorithm，and also determines weight of characteristic attributes based on the degree of correlation between characteristic attributes and target attribute to establish the intrinsic relation between the degree of correlation and weight.The simulation experiment shows that compared with traditional KNN algorithm the improved KNN algorithm based on entropy method raises the rate of accuracy enormously in classification，meanwhile it also maintains efficiency of classification.

Key words: K-Nearest Neighbor（KNN） Algorithm, entropy weight, attributes reduction, classification

摘要： 维度灾难直接影响到K最临近算法（KNN）的效率和准确率，将信息论中的信息熵理论与KNN算法结合起来，用信息熵理论进行属性约简，并根据特征属性与分类的相关度来确定各属性的权限，从而建立相关度与权重的内在联系。仿真实验表明，与传统的KNN相比，基于熵权的KNN改进方法在保持分类效率的情况下，使分类器的准确率得到了极大的提高。

关键词: K最邻近算法, 熵权, 属性约简, 分类