基于小生境及模糊先验提取的自适应免疫网络

Adaptive immune network based on niche technology and fuzzy prior extraction

为了提高人工免疫网络的数据处理能力,在aiNet的基础上提出一种动态自适应免疫网络算法。首先调用模糊C均值对(部分)初始数据进行聚类,从聚类结果的各类数据中分别选取若干个数据作为免疫网络初始抗体集合,以获取待分类数据的初始分布信息,避免因随机获取初始抗体集合而存在的盲目性;引入基于小生境的评价函数,以有效衡量抗体与抗原之间的亲和度,避免进化过程中的近亲繁殖;在网络进化过程中,对克隆抗体的选取及网络抑制的阈值进行了动态控制,使其随网络的进化而动态改变,从而提高系统的动态自适应性。通过实验测试对比分析了算法的分类精度和网络收敛速度,结果表明,该算法对数据分类具有较高的准确性,同时大大提高了网络的收敛速度。

To improve the data processing capabilities of the artificial immune network, a dynamic adaptive immune network was proposed based on aiNet. The fuzzy C-means was used to cluster part of the original data, on which a certain data of each category were selected respectively to be the initial antibodies. Thus the distribution information of the original data was obtained, which could reduce the blindness led by initial antibodies selection randomly. To avoid the inbreeding during the immune evolution processing, the fitness function based on niche was introduced which was used to effectively measure the affinity between antibody and antigen. The selection of clone antibody and the threshold of network suppression were dynamically controlled in the process of network evolution, which made them change dynamically along with the evolution of network The adaptability of the proposed algorithm was im- proved. The experiments were used to analyze and testify the efficiency and accuracy of the proposed algorithm, and the results showed that the proposed algorithm had higher classification accuracy, and the network convergence rate was greatly increased comparing to several other methods.