一种人工免疫的自适应谱聚类算法

An Adaptive Spectral Clustering Algorithm Based on Artificial Immune

聚类分组数的自动确定是谱聚类算法中一个亟待解决的问题.针对谱聚类算法聚类分组数的获取问题,提出一种基于人工免疫的自适应谱聚类算法.该算法通过模拟抗体的克隆选择机制和免疫系统的初次免疫应答、二次免疫应答机制,实现了数据样本聚类分组数的自动调整,解决了聚类算法需要人工输入聚类分组数的弊端.并分别在线性模拟数据、非凸模拟数据和UCI数据集上验证了算法的可行性、算法在非凸数据集上的优势以及算法的有效性.实验结果表明该算法可以自动获取正确的聚类分组数,提高聚类效果,减少达到全局最优解时的迭代次数,具有较高的稳定性.

To ascertain the number of clustering groups automatically is a problem which is to be solved urgently in the spectral clustering algorithm. According to the number of clustering groups problem of the spectral clustering algorithm, this paper presents an adaptive spectral clustering based on artificial immune. The algorithm simulates antibodies＇ clone selection mechanism, primary response, secondary immune response mechanism. Finally the clustering numbers of sample data can be adjusted automatically and the algorithm solves the malpractice that group number is inputted by hand. The paper validates the feasibility, validity of the algorithm on linear simulation data, non convex data and UCI datasets. The experiment results show that the algorithm can ascertain the number of clustering groups automatically, the clustering results are ultimately improved, and the iteration number is reduced when it reaches the global optimal solution. The algorithm has a higher stability.