基于PSO算法的离群点检测算法.docVIP

PSO高维离群点检测算法研究 刘培奇，孙靖，段中兴 （西安建筑科技大学信息与控制工程学院 西安，710055） 摘 要：因离群点检测在高维环境中易陷入“维灾难”问题，使离群点检测算法效率较低。本文经过对PSO算法和K均值算法的研究，提出一种基于改进粒子群优化算法的离群点检测算法，重点解决高维环境下离群点挖掘效率偏低的问题。在新算法中，通过改进进化权重和引入随机迁跃函数，改进了粒子群优化算法，并在算法中融合了k均值算法。新算法能够充分发挥粒子群优化算法全局搜索的优势，并具有k均值算法快速收敛的特点。可避免粒子群优化算法的早熟、减小确定k均值算法聚类中心的计算量等问题。实验表明，该算法在高维环境下可快速有效的计算出离群数据的离群支持度，有较好的挖掘效率、准确率和实用性。 关键字：数据挖掘；离群点检测；高维数据；PSO算法；k均值算法 中图分类号：TP311 文献标识码：A 文章编号： Research of Particle swarm optimizer high dimension outliers detection algorithms LIU Pei-qi, SUN Jing, DUAN Zhong-xing (School of Information and Control Engineering, Xi’an University of Architecture and Technology, Xi’an 710055, China) 【Abstract】n high dimensional data environment outlier detection algorithms are easy to fall into dimension disaster problem, So high dimensional data environment makes the outlier detection algorithm efficiency low. Through to PSO algorithm and k-means algorithm of research, a kind of outlier detection algorithm based on the improved particle swarm optimization algorithm is proposed in the paper here. And the paper focuses to solve the problem that mining algorithms had low efficiency in high dimensional environment. New algorithm using the weight of evolutionary algorithm and random moving step function improve particle swarm optimization algorithm, and new algorithm unite k-means algorithm in it. At the same time, the algorithm uses the global search of particle swarm optimization algorithm and the rapid convergence of k-means algorithm advantages. So that, the algorithms can avoid the premature convergence of the particle swarm optimization algorithm and reduce the amount of calculation clustering center for the k-means algorithm. Experiments show that this algorithm, in high dimensional environment, can be computing outliers the support degree of outlier data, which making outlier detection efficiently, accurately and practicably. 【Key wordsData mining; Outlier detection; High dimensional data; Particle swarm optimization; K-mean