引导函数支配的进化模糊聚类算法

建立了多维属性样本的模糊聚类目标函数.构建了引导进化算法收敛的指数函数曲线模型,给出了模型的参数计算方法.设计了一种具有全局变异和局部变异算子的进化模糊聚类算法,根据全局变异前后个体适应度值和分量值的变化趋势,实现定向变异,并给出了算法的种群进化策略.选择文本分类和点聚类计算实例,实验表明,设计的引导函数是有效的.进化模糊聚类算法具有较强的局部寻优能力,在收敛速度和聚类精度方面优于比较的遗传模糊C-均值聚类等算法.

Evolutionary fuzzy clustering algorithm dominated by guided function

An objective function of fuzzy clustering is constructed for sample clustering with multi-dimensional properties. It establishes the curve model of exponential function to guide the evolutionary algorithm convergence, and gives the calculated method of model parameters. An algorithm of evolutionary fuzzy clustering (EFC) with global mutation operator and local mutation operator is designed, which local mutation operator can achieve directional mutation according to the trend of fitness and component value of individual before and after global mutation. And then, it gives the strategy of population evolution of EFC. The experiments have been done for text classification and points clustering in Matlab, the results show that the EFC has good local search capabilities, its convergence speed and accuracy of solution guided by the valid function is better than the compared algorithms, such as algorithm of genetic fuzzy C-means clustering, etc.