一种用于数据挖掘的差异粒子群算法

聚类分析是数据挖掘的一种重要方法。本文受鸟群聚集行为的启发,提出了一种新的聚类方法——差异粒子群聚类算法(DPSC)。DPSC算法将样本数据转化为具有行为能力的粒子,通过促使不同类别的粒子相互分离,而同类的粒子高度聚集,最终在粒子的运动中涌现出整个数据群体的内部组织结构,从而完成对样本数据的聚类。在三个标准数据集和六个人工复杂数据集上同K-Means、PSO和PSO+K-Means算法进行了比较,结果显示DPSC算法的聚类性能更优异。

A Dissonant Particle Swarm Algorithm for Data Mining

Clustering analysis is an important tool of data mining. Enlightened by the collection behaviors of a flock of birds, a new data clustering algorithm  named Discriminating Dissonant Particle Swarm Clustering (DPSC) is presented. The DPSC algorithm changes data samples into a dynamical particle swarm, promotes the inhomogeneous particles to separate and the congeneric particles to collect. The structural features of the complex dataset will be emerged during the movement of particles, and the result of data object clustering is therefore achieved. Through experiments implemented on three standard datasets and six artificial complex datasets, the results show that the DPSC algorithm is more effective than the K Means, PSO and PSO+K Means algorithms.