一种带混合进化机制的膜聚类算法

膜计算(也称为P系统或膜系统)是一种新颖的分布式、并行计算模型.为了处理数据聚类问题,提出了一种采用混合进化机制的膜聚类算法.它使用了一个由3个细胞组成的组织P系统,为一个待聚类的数据集发现最优的簇中心.其对象表示候选的簇中心,并且这3个细胞分别使用了3种不同的进化机制:遗传算子、速度-位移模型和差分进化机制.然而,所使用的速度-位移模型和差分进化机制是结合了这个特殊膜结构和转运机制所提出的改进版本.这种混合进化机制能够增强系统中对象的多样性和改善收敛性能.在混合进化机制和转运机制控制下,这种膜聚类算法能够确定一个数据集的良好划分.所提出的膜聚类算法在3个人工数据集和5个真实数据集上被评估,并与k-means和几种进化聚类算法进行比较.统计显著性测试建立了所提出的膜聚类算法的优势.

Membrane Clustering Algorithm with Hybrid Evolutionary Mechanisms

Membrane computing, known as P systems or membrane systems, is a novel class of distributed and parallel computing models. This paper proposes a membrane clustering algorithm using hybrid evolutionary mechanisms to address data clustering problem. It uses a tissue P system consisting of three cells to find the optimal cluster centers for a data set to be clustered. Its object is used to express candidate cluster centers, and the three cells use three different evolutionary mechanisms: genetic operators, velocity-position model and differential evolution mechanism. Particularly, the velocity-position model and differential evolution mechanism used in the process are the improved versions proposed in this paper according to the special membrane structure and communication mechanism. The hybrid evolutionary mechanisms can enhance the diversity of objects in the system and improve the convergence performance. Under the control of the hybrid evolutionary mechanisms and communication mechanism, the membrane clustering algorithm can determine a good partition for a data set. The proposed membrane clustering algorithm is evaluated on three artificial data sets and five real-life data sets and compared with k-means and several evolutionary clustering algorithms. Statistical significance tests have been performed to establish the superiority of the proposed membrane clustering algorithm.