基于并行处理的蚁群聚类算法的研究.doc

PAGE 硕士学位论文 基于并行处理的 基于并行处理的聚类蚁群算法的研究 Based on parallel processing of the ant clustering algorithm of flocking Based on parallel processing of the ant clustering algorithm of flocking 摘 要 聚类就是将数据对象划分到不同组（或簇）中，使得属于同簇内的数据对象具有相似性，而不同簇的数据对象具有相异性。聚类分析又称群分析，它是研究样品或指标分类问题的一种统计分析方法，它是由若干模式组成的，以相似性为基础，在一个聚类中的模式之间比不在同一聚类中的模式之间具有更多的相似性，是重要的数据挖掘技术。近几十年来，国内外的学术者提出了诸多的聚类算法，力图寻找最优方案。随着蚁群算法研究的兴起，人们发现采用蚁群模型进行聚类能够更加有力的解决现实问题。 本文主要先研究了业界一些蚁群算法和聚类算法，充分深入了解了有关聚类蚁群算法的基本原理和特性。而通过研究发现人工蚁群算法本质上是一个并行系统，因此，研究并行蚁群算法对于提高运算速度具有重要的意义，在归纳总结的基础上，本文了将并行算法和聚类蚁群算法相结合，提出了一种新的聚类蚁群优化算法，同时将改良后的优化算法针对传统的TSP问题、二次分配问题进行了对比，实验结果表明该算法不仅是有效的，而且其性能更加的优越。 本文提出了并行性和聚类蚁群相结合的方法，给出了一种并行蚁群算法，该算法使用并行搜索，并且采用根据目标函数值自动调整蚂蚁搜索路径和基于目标函数值的启发式信息素分配策略。为人们研究聚类提供了新思路和新途径，因此本文的研究具有一定的理论和实践意义。 关键词：并行性；聚类蚁群；优化 Abstract Clustering is dividing data object into different groups (or cluster), make belong to same cluster the data objects with the similarity. However different data objects in clusters have different attribute. Clustering analysis, also called study of analysis, it studies a statistical analysis method of sample or index classification problem. It is composed by several mode based on similarity, and in mode of cluster has more similarity than that in the different clusters between. It is an important data mining technology. In recent decades, the domestic and international academic proposed many clustering algorithms, and tries to find the optimal scheme. With the rise of ant colony algorithm, people found using ant colony optimization model clustering can solve practical problem more powerfully Firstly this paper studied some ant colony algorithm and clustering algorithm, fully understanding the basic principle of flocking and characteristics of ant clustering algorithm. And through the researches show that people ants swarm algorithm is essentially a parallel system, therefore, Studying the parallel ant colony algorithm has an important meaning to improve speed. On the basis of summarization, combining industria