网络故障诊断研究中一种优化的否定选择算法

Optimized negative selection algorithm on research of fault diagnosis of network

针对传统的否定选择算法(NSA)在网络故障诊断应用中所生成的检测器效率不高以及检测器之间重叠面积较大的问题,提出了一种基于差分进化的改进否定选择算法(DE-NSA)。该算法先采用否定选择算法随机地产生检测器,然后通过差分进化算法对所生成的检测器进行优化分布;之后利用局部离群因子(LOF)作为适应度函数来优化检测器之间的距离,避免检测器之间过大的重叠区域。通过对网络故障数据实验仿真,以检测率、误报率、测试时间等评比标准与标准的否定选择算法相比,该方法具有一定的可行性和高效性。

To solve the problem of the low efficiency and large overlapping of between the detectors in the fault diagnosis of the network based on the negative selection algorithm （ NSA）, this paper proposed a new optimizing negative selection algorithm （DE-NSA） based on the differential evolution. The algorithm employed the negative selection algorithm and the differential evolution algorithm to generate and optimize the distribution of the detectors. The proposed method also adopted the local outlier factor （LOF） as the fitness function to optimize the distance between the detectors to avoid the overlapping area of them. This paper tested the proposed method based on the simulation experiment of the fault diagnosis in the network. Comparing with the standard negative selection algorithm, the method outperformed in the terms of detection rates, false alarm rates, and the time consuming. Thus, the results illustrate the feasibility and validation of the method in the research of the fault diagnosis in the network.