摘要
在PSO融合FCM实施聚类分析中,为克服PSO迭代后期易于发生早熟这一问题,选用Chebyshev映射产生混沌序列。在粒子群初始化时,使用该映射分别初始化各粒子位置和速度,同时,在粒子群算法各次迭代运行中,使用该映射计算惯性系数,并利用适应度方差判定粒子群算法是否发生早熟。若未发生早熟,则依基于混沌惯性系数粒子群搜索最优解,当发生早熟时,则按当前粒子群迄今为止搜索到的最优位置为起点进行混沌搜索,并用搜到的最优位置替换粒子群中最差粒子位置,进而将该混沌粒子群算法同FCM算法融合完成聚类分析任务。提出一种基于Chebyshev映射的混沌粒子群融合FCM均值聚类算法。实验结果显示该算法具有较好的寻优能力并提高了样本分类精度。
In clustering analysis executed by integrating particle swarm optimisation （PSO） with fuzzy c-means （FCM）, for overcoming the problem that in later phase of PSO iteration it is prone to prematurity, we choose Chebyshev map to generate chaotic sequence. In initialisation of the PSO, this map is used to initialise the positions and velocities of each particle respectively, and at the same time to calculate the inertia coefficients when each order of iteration in PSO is running. Moreover, the fitness variance is used to judge whether the PSO has the prematurity happened or not, if not, the optimal solution will be searched in accordance ,~ith the chaotic inertia coefficient PSO; if does, the best position so far searched by current PSO will be used as the starting point to carry out chaotic search, and the position of the worst particle in PSO will be substituted by the best position searched. Furthermore, the chaotic-PSO is integrated with FCM to complete the task of clustering analysis, in this way we present a Chebyshev map-based clustering algorithm integrating the chaotic PSO with FCM. Experimental result indicates that the algorithm has quite good optimisation capability and enhances the accuracy of sample classification.
出处
《计算机应用与软件》
CSCD
2015年第7期255-258,共4页
Computer Applications and Software
基金
安徽省高校优秀青年人才基金项目(2012SQRL154)
安徽省高等学校自然科学研究项目(KJ2010B146)
