中文摘要：

现有的自适应亲和传播聚类存在聚类时间长、精度低的缺点,提出了一种结合半监督的改进自适应亲和传播聚类（SAAP）。它首先利用半监督学习更新相似度矩阵,而后在亲和传播聚类的基础上,通过基于二分法判断实现自适应搜索有效聚类数空间,最后由加权评价函数确定最佳聚类。经实验证明,SAAP算法可以更快速地扫描有效聚类空间,并能够得到较小的错分率和较高的有效性评价。

英文摘要：

The existing adaptive affinity propagation clustering has some shortcoming,such as long runtime and low accuracy. This paper proposed an improved adaptive affinity propagation clustering based on semi-supervised learning （ SAAP）. It first updated the similarity matrix by semi-supervised learning,then scaned adaptively the effective clustersing space based on Affinity propagation clustering by the dichotomy judge,and finally determined the best clustering by the weighted evaluation function. The experiments show,SAAP algorithm can more quickly scan effective clustering space,and can be smaller misclassification rate and a higher effectiveness evaluation index.