面向多类别模式分类问题的新型阴性选择算法

Novel negative selection algorithm for multi-class pattern classification problems

针对阴性选择算法缺乏高效的分类器生成机制和"过拟合"抑制机制的缺陷,提出了一种面向多类别模式分类的阴性选择算法CS-NSA。通过引入克隆选择机制,根据分类器的分类效果和刺激度对其进行自适应学习;针对多类别模式分类的"过拟合"问题,引入了检测器集合的修剪机制,增强了检测器的分类推广能力。对比实验结果证明:与著名的人工免疫分类器AIRS相比,CS-NSA体现出更高的正确识别率。

A negative selection algorithm for multi-class pattern classification problems named CS-NSA was proposed. The algorithm used clonal selection mechanism to implement self-adaptive learning of detectors and adopted detector trimming mechanism to tackle the over-fitting problem in multi-class classification. This mechanism enhanced the generalization capability of the detectors. The results of comparative experiments show that the proposed algorithm exhibits higher classifying accuracy than that of AIRS, a famous artificial immune classifier.