基于自适应权重粗糙K均值聚类算法.pdfVIP

38 6 Vol. 38 No. 6 2011 6 Computer Science June 2011 K ( 同济大学电子与信息工程学院 上海201804) ( 同济大学嵌入式系统与服务计算教育部重点实验室 上海201804) ( 家高性能计算机工程中心同济分中心 上海201804) 原有Rough Kmeans 算法中类的上下近似采用固定经验权重, 其科学性值得商榷, 针对这 一 题, 设计了 一种基于自适应权重的粗糙 K 均值聚类算法基于自适应权重的粗糙聚类算法在每 一次迭代过程中,根据当前的数 据划分状态, 动态计算每 个样本对于类的权重, 降低了原有算法对初始权重的依赖此外, 该算法采用近似集合中的 高斯距离比例来表现样本权重, 从而可以在多种数据分布上得到更精确的聚类结果实验结果表明, 基于自适应权重 的粗糙K 均值算法是一种较优的聚类算法 聚类, 粗糙集, 粗糙 K 均值, 自适应权重 T P391 A Rough Kmeans Clustering Based on Selfadaptive Weights ZHOU Yang MIAO Duoqian YUE Xiaodong ( The School of Electronics and Inform ation Engineering, Tongji University, Shanghai 201804, China) ( T he Key Laboratory of Embedded System and Service Computing Ministry of Education, Shanghai 201804, China) ( T ongji Branch, National Engineering T echnology Center of High Performance Computer, Shanghai 201804, China) Abstract The fixed w eights are adopted in t he tradit ional rough Kmeans algorithm to represent the different approxi mations of t he clusters, but it is alw ays difficult to predefine the optimal weights w ith little priori know ledge before clustering. Therefore, an improved rough Kmeans algorithm based on selfadaptive w eights was proposed in this paper. The new method computes the w eight s for every data according t o the current clustering state and no more does rely on the initial w eights. Furthermore, the selfadaptive w eight s are obtained from the Gaussian distance ration in cluster ap proximat ion,w hich can lead to the more accurate clustering results. T he experiments indicate that the rough Kmeans based on selfadapt ive w eights is an effect ive rough c