不同的距离测量方法对人工免疫识别系统的性能影响*

人工免疫识别系统AIRS（Artificial Immune Recognition System）是著名的免疫网络分类器，被成功地应用到大量的分类问题，表现出了良好的性能。为了分析不同的距离测量方法对AIRS的性能影响， 采用三种距离测量方法实现AIRS，这三种方法分别是Euclidean距离、Manhattan距离和RBF核空间距离，并将三种用不同距离测量方法实现的AIRS算法应用于Iris，Heart和Wine数据集的分类测试。所获得的三组数据集分类的准确率和抗体规模进行了相互比较，结果表明采用Manhattan距离AIRS算法获得了对Iris和Heart的最高分类准确率，而采用核空间距离，算法获得了对Wine的最高分类准确率。从抗体群体规模来看，采用核空间距离则能获得最小的抗体群体。从性能比较可知，不同的距离测量方法对AIRS算法的分类性能较大的影响。

Effect of different distance measure methods on performance of artificial immune recognition system

Artificial Immune Recognition System (AIRS) is a famous immune network classifier, which has been applied to many of classification problems and achieved high classification performance. In order to analyze the effect of different distance measure methods on the AIRS classification performance, we implemented AIRS with three different distance measure methods, i.e., Euclidean distance, Manhattan distance and RBF based kernel space distance, the classifiers were used to Iris, Hear and Wine problems. The obtained three groups of results were compared with each other with regard to the classification accuracy and memory cells. The results showed that the AIRS implemented with Manhattan distance measure method reached the highest accuracies for Iris and Heart among the three versions of AIRS and the highest accuracy reached for Wine was the AIRS with kernel space distance measure method used. Moreover, the most compact memory population was produced by the AIRS with kernel space distance measure method. It can be seen from the comparisons that different distance measure methods give effect on the performance of AIRS to some extent.