摘要
电力行业信息化建设在近几年得到飞速发展的同时,其数据却得不到很好地利用以支持决策。聚类分析是数据挖掘领域的重要分支,将其应用于电力行业,提升电力企业的市场竞争力。本文主要研究聚类分析算法在电力营销客户细分领域的应用。由于在细分客户的时候,并不能明确指出某个客户一定属于或不属于某一客户群体,恰恰相反一个客户样本可以属于不同的客户群,而模糊聚类算法由于引入隶属度的概念能更好地体现客户的这一特征,所以本文采用模糊c均值(FCM)聚类算法来对电力客户进行细分。在验证FCM聚类算法有效性时本文采用的是基于Xie-Beni有效性的方法,并对该方法进行了修改,另外又在FCM算法的循环之后添加了去除空簇的步骤,提高了算法的效率。在实验阶段,通过选用不同的参数细分来得到不同的细分结果,并对结果进行分析,验证了FCM算法的有效性;最后将改进后的FCM算法应用于电力营销决策支持系统中对客户进行细分。
Informationization construction in electric power has been developed in recent years. Meanwhile,data of power plants cannot be well utilized to support decision.Clusteringanalysis, as an important embranchment of data mining,should be applied to electricpower industry, which will enhance market competitiveness of power plants.Along with restructuring of electric power system,the paces of the power supply enterprise undertaking marketing go faster. The thesis mainly researches on the application of clustering analysis in customer of electric marketing segmentation. It is hard to clearly say that a customer belongs to or does not belong to a particular customer group while segmenting customers. On the contrary, a customer can belong to different groups. Fuzzy c-means(FCM) clustering algorithm can commendably reflect the characteristics of customers for the introduction of membership concept, so fuzzy c-means(FCM) clustering algorithm is adopted here. The method used for validating the FCM validity is based on the Xie-Beni validity, the thesis modifies the Xie-Beni validity and a step of eliminating empty clusters is added after the FCM loop to improve the efficiency of the algorithm. In the experiment phase, we use then the validity of the algorithm is validated by analyzing these results. Finally,the improved FCM algorithm used in electric power marketing decision support system for customer analysis.
引文
[1]Jiawei Han,Micheline Kamber,Data Mining Concepts and Techniques,北京:机械工业出版社,2001(11)
[2]徐春玲,姜雨,郑孝安,基于应用服务器的三层体系结构的电力营销系统,电网技术,2002,26(7):71-73
[3]刘玉梅,王燕,陈澍,客户关系管理系统数据仓库的建立,信息技术,2003,27(1):20-22;
[4]高人伯、陈文伟,数据仓库与数据开采相结合的决策支持新技术,计算机世界专题综述,1997年6月30日;
[5]石丽、佟贺、李洪凯、李坚,供电预测数据仓库的建设,电力系统及其自动,2003,27(13):75-77;
[6]肖刚.,Java Web服务器应用程序设计,电子工业出版社,2001;
[7]陈斌蒋绍忠,浅谈客户关系管理系统(CRM)中的数据处理,企业经济,2002,(11):81-82; 841-847.
[8]高新波著.模糊聚类分析及其应用.西安电子科技大学出版社.2004,49-62
[9]张敏,于剑.基于划分的模糊聚类算法.软件学报.Vol.15,N.o6,200.4
[10]J.C.Bezdek.Pattenrercongitionwithuzfzyobjeetiveufnetionalgorithlns.New York:Plenup Press 1981.
[11]Yi Shen,Hong Shi,and JinaQiu Zhang.Imporvement and OPtimization of a Fuzzy C-Menas Clustering Algorithm.EIEE Instrumentation and Measurement Technology Coneference Budpaest,Hungary,May,2001,21-23.
[12]Carl G.Looney. A Fuzzy Clustering and Fuzzy Merging Algoritmh.Computer Science Department/171,University of Nevada,Redo,NV89557.1999.
[13]石洪波,于剑,黄厚宽.一种有效的FCM算法的实现方式.铁道学报,vol.25,No.1,February,2003.
[14]刘宜平,沈毅,刘志言一种FCM聚类算法的改进与优化.系统工程与电技术,Vol.22,No.4,2000.
[15]Haojun Sun,Shengrui Wang,Qingshan Jiang.FCM-Based Model Selection Algorithms for Determining the Number of Clusters. Pattern Recognition No.37,2004,2027-2037.
[16]Carl G.Looney. Interactive clustering and merging with a new fuzzy expected value.ComPuter Science Department/171 , University of Nevada , Redo , NV89557 , USA.PatternRecongnition.2001.
[17]宫改云,高新波,伍忠东.FCM聚类算法中模糊加权指数m的优选方法.模糊系统与数学.Vol.19,No.l,2005,143-148.
[18]Xizhao Wang,Yadong Wang,Lijuna Wang. Improving fuzzy c-means clustering based on featuer-weight learning.Pattern Recognition Letters25,2004,1123-1132.
[19]A Novel Implementation Method of the FCM.SHI Hong-Bo,YU Jian,and HUANG Hou-Kuan.ICSP’02Proceeedings,?2002IEEE.
[20]高新波,裴继红,谢维信.模糊c-均值聚类算法中加权指数m的研究.电子报,Vol.28,No.4,April,2000.
[21]于剑,程乾生.关于FCM算法中的权重指数m的一点标记.电子学报.VOl.31,No.3,Mar,2003.
[22]肖刚.,Java Web服务器应用程序设计,电子工业出版社,2001;
[23]胡健,电力市场营销管理[M],北京:中国电力出版社,2000;
[24]刘慧,浅谈电力营销中电费与电价的分析方法,华中电业,2002,1:54-56;
[25]廖若雪,JSP高级编程,机械工业出版社,2001;
