Abstract: Aiming at the existing problems of negative selection algorithm, such as low detection rate and large detector set,?this paper proposes a new real-valued negative selection algorithm, which considers non-self information and twice move point, called NTMV-detector. This algorithm generates candidate detector center based on the training of non-self and the random method. The main idea of twice move point is: if the candidate detector center matches with mature detectors, it will move out of mature detector set, instead of giving up the detector center. Then, the algorithm adjusts the location of the detector center and determines detector radius through two closest self-data to the detector center. The experiments prove that this algorithm can effectively improve diagnosis rate and reduce misdiagnosis rate.

Key words: negative selection algorithm, detector, real-value, non-self information

摘要： 针对现有的否定选择算法存在检测率较低，检测器集合过大等问题，提出了一种结合非自体信息和二次移动的实值否定选择算法（NTMV-detector）。该算法基于训练集中的非自体和随机的方法生成候选检测器中心。二次移动的主要思路是：如果候选检测器中心与成熟检测器匹配，把它移出成熟检测器集；然后通过离候选检测器中心最近的两个自体来微调检测器的位置，确定检测器半径。实验证明，该方法可以有效地提高疾病诊断的诊断率，降低误诊率。

关键词: 否定选择算法, 检测器, 实值, 非自体信息