一种基于改进神经网络的高效模糊聚类算法  

Highly effective fuzzy clustering algorithm based on improved network

作　　者：陈秀敏[1] 邹开其[1] 闫忠文[2] 祝美宁[2] 付长青[1] 杨艳萍[2] 阎丹丹 

机构地区：[1]大连大学信息工程学院,辽宁大连116622 [2]河北科技师范学院计算机系,河北秦皇岛066600 [3]昌黎职教中心,河北秦皇岛066600

出　　处：《计算机应用》2008年第5期1190-1193,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(60573072)

摘　　要：针对利用自组织特征映射(SOFM)神经网络进行模糊聚类时出现的一些问题,提出改进结构的神经网络,采用自适应的聚类初值,能够实现高维数据和任意形状族的聚类,与具有同样聚类效果的其他算法相比,具有较低的时间复杂度。仿真实验结果表明,该聚类算法比单个的神经网络聚类算法和同类其他算法更有效。In order to solve the problems in fuzzy clustering by using Self-Organizing Feature Map （SOFM） network, this paper introduced an improved structural self-organizing feature map network and adopted self-adapting initial condition. It can handle the clustering problem of high dimensional data and the clusters with arbitrary shapes. Compared with other algorithms with the same clustering effect, it has lower clustering time complexity. Experiments indicate this algorithm has better clustering effect compared to single SOFM network and other kin algorithms.

关 键 词：聚类 自组织特征映射 拓扑相似度 自适应 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]