一种新的K—means最佳聚类数确定方法

在传统的K-means算法中，聚类数K是随机给定的，K值选取不合理会造成K—meall$算法陷入局部最优。针对这个缺点，提出一种新的K—means聚类数确定方法，根据聚类算法中类内相似度最大差异度最小和类问差异度最大相似度最小的基本原则．提出距离评价函数作为最佳聚类数的检验函数，建立相应的数学模型，并通过实例结果进一步验证新算法的有效性。

A New Method of Determining the Optimal Class Number for K-means

In traditional K-means algorithm, cluster number K is given at random ,not reasonable K value will cause K-means to local optimal. Considering this defection, proposes a new method to de- termine the class number. According to the basic principles of clustering algorithm, that the within-class similarity is maximum and the within-class difference is least,the inter-class dif- ference is maximum and the inter-class similarity is least, recommends a distance cost of func- tion to confirm the optimal class number, sets up a corresponding math model, and validates the effectiveness of the new algorithm by example results