基于核聚类的近邻人工免疫网络算法研究

针对经典人工免疫网络(aiNet)及改进算法中存在的运算时间长、结构复杂等问题,提出了一种改进的核聚类近邻人工免疫网络算法(KN-aiNet)。算法在aiNet的改进算法——近邻aiNet结构的基础上,以抗体数据为核心利用量子能级思想聚类,并重定义了生成抗体策略,采用区域生长法搜索拥挤距离,采用基于核函数的亲和度等方法来提高算法的聚类效果和降低算法的运算时间。聚类实验结果表明,KN-aiNet算法的聚类准确率较经典aiNet算法及近邻aiNet算法分别提高了11.53%和4.56%,而算法的运算时间较经典aiNet算法及近邻ai-Net算法分别下降了0.503 s和0.823 s。

Study in neighbor artificial immune net work based on kernel clustering

To solve the issue of high time complexity and algorithm structure of aiNet, this paper proposed a modified aiNet algorithm:KN-aiNet. Based on the structure of neighbor aiNet, the algorithm took antibody data as core, then clustered data around the core, used energy level as affinity. The algorithm redefined the creation of antibody, took region growing to match distance, took kernel function to improve clustering effect and reduce time complexity. The simulation proves the clustering accuracy of KN-aiNet improves 11. 5% compared with aiNet, 4. 56% compared with neighbor aiNet, and time complexity declines 0. 503 s compared with aiNet, 0. 823 s compared with neighbor aiNet.