中文摘要：

传统的否定选择算法无法有效识别落入到低维子空间的样本，导致算法在高维空间检测性能不佳。为此，提出了面向子空间的否定选择算法（subspace—orientednegativeselectionalgorithm，SONSA）。在训练常规检测器的基础上，SONSA将搜索样本分布密度较高的低维子空间以进一步训练面向子空间的检测器，从而提高算法对低维子空间内样本的识别能力。实验结果表明，在标准数据集Haberman’SSurvival（三维）与BreastCancerWisconsin（九维）上，相对于经典的V-Detector算法以及采用PCA降维的V—Detector算法，SONSA能在误报率相似的情况下显著地提高检测率。

英文摘要：

Traditional negative selection algorithm cannot distinguish these samples effectively, it causes the negative selection algorithm has a poor performance in high-dimensions features space. To deal this situation, this paper proposed a subspace-ori- ented negative selection algorithm （SONSA）. Besides training conventional detectors, SONSA would search out the subspace where the samples might distribute densely, then generated the subspace-oriented detectors for covering the aimed subspace as much as possible, thus improved detection rate of algorithm. The experimental result shows that, on the Haberman＇ s Survival dataset （31）） and Breast Cancer Wisconsin dataset （gD）, compared to the classical V-Detector algorithm and PCA V-Detec- tor algorithm, SONSA can reach the higher detection rate with the similar false alarm rate.