蚁群聚类组合的改进算法  

Primary Research on Improved Algorithm of Ant Colony Clustering Combination

作　　者：马春英[1] 曹安得[2] 周允征[3] 

机构地区：[1]沈阳大学工商管理学院,辽宁沈阳110044 [2]沈阳大学科技中心,辽宁沈阳110044 [3]沈阳师范大学财务处,辽宁沈阳1100034

出　　处：《沈阳建筑大学学报（自然科学版）》2011年第4期798-803,共6页Journal of Shenyang Jianzhu University：Natural Science

基　　金：国家自然科学基金项目(70473059)

摘　　要：目的研究一种基于信息熵蚁群聚类的模糊C-均值算法(即EACFCM算法),解决原标准蚁群聚类算法聚类速度慢,系统停滞,函数值收敛于某局部点等问题.方法通过理论改进、算法移植和实验证明相结合的办法,采用基于信息熵的蚁群聚类算法对数据样本进行聚类(一次聚类)以获得新的聚类中心,然后从聚类拆分、合并、孤立点处理等几个方面对基于信息熵的蚁群聚类算法进行改进,以提高前期所获得的新聚类中心的准确性,最后再利用模糊C-均值聚类算法(FCM算法)对聚类结果进行优化处理(二次聚类).结果 EACFCM算法在迭代180次后,得到的目标函数值60.245是FCM算法函数值660.135的十分之一;划分熵的函数值0.440明显小于FCM算法函数值0.491;分离度的函数值0.102明显大于FCM算法函数值0.091.结论笔者改进的EACFCM算法得到的聚类比传统的FCM算法效果好;仿真结果表明了其有效性.This paper discussed and developed an improved pheromone of ant colony clustering algorithm based on information entropy and combination of the FCM（i.e.FCM algorithm）to avoid problems such as the slow clustering of the standard ant clustering algorithm,system stagnation and function values converging to certain local points.By means of improving theories,algorithm transplantation and demonstration,the new clustering center was obtained by clustering data sample using the method of ant colony clustering algorithm based on information entropy（LF method）.The researchers then improved it from several aspects of clustering split,merger and outlier processing,in order to improve the accuracy of early new obtained clustering center.Finally,the clustering results were optimized by the fuzzy-c-means clustering algorithm（EACFCM method）.After 180 times＇ iteration,the EACFCM method gained the results that the function value 60.245 is one-tenth of the FCM algorithm function value 660.135,and the partition entropy function value 0.440 is obviously less than FCM algorithm function value 0.491,and the separation degree function values 0.102 is significantly greater than the value of FCM algorithm 0.091.The simulation results showed its validity that better effects can be obtained by improved EACFCM algorithm than by the traditional ones under the same conditions.

关 键 词：信息熵 聚类分析 蚁群聚类 FCM聚类算法 

分 类 号：X705[环境科学与工程—环境工程]