基于遗传算法的投影寻踪聚类

传统的投影寻踪聚类算法PROCLUS是一种有效的处理高维数据聚类的算法,但此算法是利用爬山法（Hill climbing）对各类中心点进行循环迭代、选取最优的过程,由于爬山法是一种局部搜索（local search）方法,得到的最优解可能仅仅是局部最优。针对上述缺陷,提出一种改进的投影寻踪聚类算法,即利用遗传算法（Genetic Algorithm）对各类中心点进行循环迭代,寻找到全局最优解。仿真实验结果证明了新算法的可行性和有效性。

Projection Pursuit Clustering Method Based on Genetic Algorithm

The traditional projection pursuit clustering algorithm PROCLUS[ 1,2] is an effective method to deal with high- dimensional data clustering. However its vital shortcoming is using Hill climbing method to search the optimal prototypes of the clusters, which is easy to run into a local optimum. An improved clustering algorithm is proposed in this paper. Using Genetic Algorithm which has good global and local search capability to search the optimal prototypes of the clusters, one will get global optimum. Simulated experiments show the feasibility and efficiency of the proposed method.