数值型和分类型混合数据的模糊K-Prototypes聚类算法

由于数据库经常同时包含数值型和分类型的属性,因此研究能够处理混合型数据的聚类算法无疑是很重要的.讨论了混合型数据的聚类问题,提出了一种模糊K-prototypes算法.该算法融合了K-means和K-modes对数值型和分类型数据的处理方法,能够处理混合类型的数据.模糊技术体现聚类的边界特征,更适合处理含有噪声和缺失数据的数据库.实验结果显示,模糊算法比相应的确定算法得到的结果准确度高.

Fuzzy K-Prototypes Algorithm for Clustering Mixed Numericand Categorical Valued Data

The capacity of dealing with mixed numeric and categorical valued data is undoubtedly important for clustering algorithms because there is usually a mixture of numeric and categorical valued attributes in real databases. The use of fuzzy techniques makes clustering algorithms robust against noise and missing values in the databases. In this paper, a fuzzy kprototypes algorithm integrating k-means and k-modes algorithm is presented and is used to mixed databases. Experiments on several real databases demonstrategythat fuzzy algorithm can get better result than the corres ponding hard algorithm.Some properries of fuzzt k-prototypes algorithm are also discussed.