自构形网状聚类算法及应用研究  

Research and application of reticulate clustering algorithm based on auto-structure

作　　者：李翠[1] 吕林涛[2] 白晓东 

机构地区：[1]西安财经学院信息学院,陕西西安710100 [2]西安理工大学计算机学院,陕西西安710048 [3]横河西仪有限公司系统部,陕西西安710075

出　　处：《计算机工程与设计》2008年第13期3448-3450,3453,共4页Computer Engineering and Design

基　　金：陕西省自然科学基金项目(2007F52);陕西省教育厅专项科研基金项目(07JK339)

摘　　要：报告了聚类分析技术的现状,通过最小化目标函数的分析研究,提出了自构形网状聚类算法,以减少样本几何形状对聚类结果的影响,并给出了网状聚类的结构图。阐述与实现了其簇心全连通和簇心连通两种模型,分析了自构形网状聚类目标函数的组成,同时以CENTUM CS3000R3和VisualC++6.0为平台的故障检测系统为实例,根据采集数据的输入构造网状聚类进行验证,实验结果表明其聚类效果很好,验证了该算法的有效性,具有较高的理论及应用价值。The actuality of clustering is reported, based on the analysis and study of minimizing objective function, reticulate clustering algorithm based auto-structure is proposed, in order to reduce the infection of stylebook figure to clustering result, the structure graph of reticulate clustering is presented. And fully connected cluster heart model and connected cluster heart model are expounded and implemented, the composing of reticulate clustering＇s objective function is analyzed, and the failure testing system based on CENTUM CS3000 R3 and Visual C＋＋ 6.0 as its application, reticulate clustering based on collected data is validated, the experimental result shows that cluster effect is very good, this algorithm is proved to be highly effective, so it is very valuable in practice and academic study.

关 键 词：最小化 自构形 聚类 DCS OPC 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]