关键词: 人工免疫识别系统, 混合核函数, 适应度, 复杂数据

Abstract: In order to improve the nonlinearity of Artificial Immune Recognition System（AIRS） and optimize the classifier, an improved AIRS algorithm is proposed. The new algorithm adopts hybrid kernel function to improve its nonlinearity, moreover, the individual memory cell is evaluated by its fitness score and the cells with low fitness scores are pruned to optimize the classifier. The improved algorithm has been applied to the complex UCI datasets classification, the results have been compared with the results achieved by other classic classifiers. The comparison shows that the algorithm achieves better classification performances.

Key words: artificial immune recognition system, hybrid kernel function, fitness score, complex data