基于自适应差异演化的模糊聚类算法

A New Fuzzy Clustering Algorithm Based-on Adaptive Differential Evolution

在聚类分析中,模糊C-均值聚类(FCM)是一种广泛应用的算法,但由于它是基于梯度下降的,本质上是一种局部搜索算法,容易陷入局部极小值,且对初始值很敏感.本文提出一种基于自适应差异演化的模糊聚类算法(FCBADE),该算法利用差异演化良好的全局搜索能力,在全局范围内寻找最优解的近似解,然后由FCM算法在该近似解的周围进行局部搜索,最终得到全局最优解.同时为减少手工设置控制参数对DE算法的影响,采用自适应方式调整DE算法的控制参数.实验结果表明,该算法不仅有效克服了FCM算法易陷入局部极小值的缺点,而且明显地避免了对初始化选值敏感性的问题,也有较快的收敛速度.

Fuzzy C-means clustering （FCM） algorithm is a widely used algorithm in cluster analysis. However, as it is based on the gradient descent, FCM is essentially a local search algorithm. It is easy to fall into a local minimum, and is very sensitive to the initialization. In this paper, a new fuzzy clustering method based on an improved differential evolution algorithm was presented. First, the algorithm searches the approximate global optimal solution by the improved differential evolution, then the FCM algorithm is used for search in the optimal solution surrounding approximate solution. At the same time, an improvement is presented for reduce the impact of manual set parameters for DE algorithm. Experimental results shown that the proposed algorithm not only avoids the local optima and is robust to initialization, but also increases the convergence speed.