一种基于禁忌邻域的免疫网络分类算法

Atabooed neighborhood based immune network classification algorithm

为了改善免疫分类算法中记忆细胞确定机制和亲和度表示机制,提出了一种免疫网络分类算法。该算法利用训练抗原的禁忌邻域来指导记忆细胞的生成,采用高斯径向基函数来表示抗体-抗原之间的亲和度。算法被应用于标准数据集的分类,分类结果表明,引入核函数强化了对问题空间的搜索,而使用禁忌邻域来指导记忆细胞生成则对大多数问题有较大的优化作用。这些结果说明,该算法是一种性能良好的分类算法,具有潜在的应用能力。

In order to improve memory cell determination scheme and the affinity representation scheme in the immune classification algorithm, an immune network based classifica- tion algorithm is proposed. The algorithm adopts the training antigen＇s tabooed neighborhood to guide the memory cells determination, and uses Gauss radial basis function to represent the affinity between antibody-antigen. The algorithm is used for standard datasets classification, the classification performance shows that the search space is enlarged by introducing kernel function, and the taboo neighborhood based memory cell determination has significant optimal effect on majority problems. The results indicate that our algorithm is a high-performarme classification algorithm and has potential ability for application.