改进粒子群优化算法及其在聚类分析中应用

针对标准粒子群优化算法初期收敛速度快,后期容易陷入早熟收敛,局部寻优,全局搜索能力差等缺点,提出了一种新的鱼群-粒子群优化算法(AF-PSO)。引入拥挤因子δ和马尔可夫链,将鱼群算法加入到粒子群优化算法中,通过计算拥挤因子,来更新速度模型,使其在觅食,聚群,追尾,随机4种模态下进行切换。仿真结果表明了所提出的AF-PSO算法的综合性能优于其他改进的PSO算法。为进一步说明算法的实用性,将AF-PSO算法成功应用于输油管道泄露数据的聚类分析上。实验结果表明基于AF-PSO的K-means算法性能是优于其他聚类算法。

An Improved Particle Swarm Optimization Algorithm and Its Application in Clustering analysis

In this paper, a novel artificial fish swarm particle swarm optimization algorithm (AF-PSO) is proposed corresponding to the shortcomings of the standard particle swarm algorithm including the fast convergence speed in the initial stage, the easiness to fall into premature convergence in the late, the local optimization and the poor ability to global search. This paper firstly introduces the crowding factorδ and the Markov chain, and then adds the artificial fish swarm algorithm to the particle swarm optimization algorithm. By calculating the crowding factor, the velocity model is updated to switch among four modes: foraging, clustering, following and random. The simulation results show that the proposed AF-PSO algorithm has better performance compared with other improved PSO algorithms in synthesis. To further illustrate the application potential, the AF-PSO algorithm is successfully applied to the clustering analysis of oil pipeline leakage data. Experiment results demonstrate that the performance of the AF-PSO based K-means method is better than other clustering algorithms.