Abstract: To solve large numbers of computations in the problem of large-scale data clustering，a particle swarm optimization of fuzzy systems in parallel k-means clustering algorithm is proposed to deal with this problem.The method adjusts dynamically inertia weight and acceleration factor of particle swarm optimization with fuzzy rules，the problems of particle mobility loss and the end of evolution can be dealt with successfully.the algorithm maintains individual diversity and solves the premature convergence problem.Task parallelization and partially asynchronous communication of the algorithm are employed to decrease computing time.The simulation experiments indicate the algorithm helps increase computing speed and improve the clustering quality.

Key words: parallel clustering, fuzzy systems particle swarm optimization, task parallelism, asynchronous communication

摘要： 为了解决大规模的数据聚类问题时需要的大量计算，提出了一种模糊系统的微粒群优化并行k-means聚类算法。该方法利用模糊规则，动态地调整微粒群惯性权重和加速因子，克服群体逐渐失去迁移性而停止进化的问题，保证群体多样性而避免陷入局部极小值。采用任务并行和部分异步通信模式，降低计算时间。实验结果表明，该算法在并行机群上运行时，加快了聚类算法的计算速度，提高了聚类质量。

关键词: 并行聚类, 模糊系统微粒群优化, 任务并行, 异步通信