一种改进的PSO-Means聚类优化算法

An Improved PSO-Means Clustering Algorithm Based on Kernel Methods

针对粒子群优化算法在线性不可分情况下不能找到合适的聚类初始质心和正确的聚类个数的缺点,提出引入核方法,对基于粒子群算法的K均值聚类(PSO-Means)算法进行改进。利用核方法把数据映射到高维空间,在高维空间中使用粒子群算法找出所应聚的类,最后利用核空间中的聚类算法对数据进行聚类。通过实验,验证了该算法在线性不可分的情况下可以较好的运行,在很大程度上提高了聚类的效果。

Since the particle swarm optimization algorithm could not find the appropriate original centroid and the accurate cluster number under linear inseparable circumstance,an improved K-means algorithm based on particle swarm optimization was given.Kernel method was used to map the data to higher-dimensional space and particle swarm optimization was used to seek out the category that should be clustered.Then the data was clustered by the cluster algorithm.The experiment demonstrates that the new method runs well under linear inseparable situation and thus greatly increases the effect of clustering.