蚁群聚类算法的并行化设计与实现

Parallel Design and Implementation of Ant Colony Clustering Algorithm

蚁群聚类是一种有效的聚类方法,已在数据分析等领域获得广泛应用。MPI并行计算提供高效的数据处理方案,研究蚁群聚类算法的并行化是目前具有挑战性的研究课题。首先介绍了基于传统编程模型的解决TSP问题的蚁群优化算法,以及蚁群优化算法和K-means结合的聚类方法,描述了它们的基本原理和实现过程。然后,对基于传统编程模型的聚类算法进行MPI并行化改进,实现了基于MPI并行计算的蚁群聚类算法。最后,分别采用Iris、Wine、Zoo3个UCI数据集和Reuter-21578文本数据集进行多次测试,对基于传统编程模型的聚类算法和基于MPI并行计算的聚类算法进行性能和效率上的比较,得出基于MPI并行计算的聚类算法更优的结论。

Ant colony clustering is an effective clustering method, and has widely been applied in data analysis domain. MPI parallel computing provides benefits to handle with dataset efficiently. It is a challenge research topic to parallelize ant colony clustering algorithm. The paper firstly introduces the ant colony optimization algorithm to solve the TSP problem based on traditional programming model and the combining of ant colony optimization and K-means clustering algorithm, and describes their basic principles and achievement process. After that, the paper improves the clustering algorithm based on traditional programming model to MPI parallel algorithm. Finally, the paper compares the performance and efficiency between the clustering algorithm based on traditional programming model and MPI parallel computing on Iris, Wine, Zoo and Reuter-21578 dataset respectively. The experiments show that the clustering algorithm based on MPI parallel computing is better.