基于免疫克隆算法的多目标聚类方法

Multi-objective Clustering Method Based on Immune Clonal Algorithm

  • 摘要: 在分析k均值聚类和免疫进化聚类不足的基础上,提出一种基于Parzen密度估计的多目标免疫克隆聚类方法.该算法针对多目标免疫克隆算法中克隆规模难以确定的问题,根据密度聚类的思想,引入核密度估计,根据密度和进化代数确定各抗体的克隆规模,使用混沌变异增加抗体多样性.最后通过TOPSIS(technique for order preference by similarity to an ideal solution)方法进行抗体选择.人工以及UCI(universal chess interface)数据集上的仿真实验表明,该方法可以有效地提高算法速度,得到较好的聚类结果.

     

    Abstract: A multi-objective immune clonal clustering method based on Parzen density estimation is proposed after analyzing disadvantages of k-means and immune evolutionary clustering. Aiming at the problem that clonal scale is hard to be determined in multi-objective immune clonal algorithm, a method based on density clustering is developed by using kernel density estimation. And the clonal scale of antibody is determined by density and generation. Chaotic mutation is introduced into this method to increase antibody diversity. Finally, TOPSIS (technique for order preference by similarity to an ideal solution) is used to choose antibodies. The simulation experiments on artificial and UCI (universal chess interface) data sets show that this algorithm has a higher speed and better clustering result.

     

/

返回文章
返回