中文摘要：

K—prototypes算法是处理混合数据的主要聚类算法，大部分针对混合型数据的聚类算法都是选择数据集中的一部分数据作为聚类对象，而忽略了这类数据的特殊性与整体性，为了改进了数据的距离衡量，文中提出了一种新的聚类方法，该方法采用信息熵作为属性的权值，进行高精度和更加稳定的聚类，最后通过Matlab编程实现，采用uci数据集中credit等数据集进行仿真实验，证明改进算法是正确和有效的．

英文摘要：

K-Prototypes algorithm is the main clustering algorithm for processing mixed data. Since most clustering algorithms for mixed data choose parts of the data set as clustering objects, and ignore the particularity and globality of the data, a new data distance measure is improved in this paper. A new clustering method is proposed using the entropy as the attribute weights for more accurate and more stable clustering. Finally, uci data set in Matlab is used and the algorithm is proved correct and effective.