小生境粒子群支持向量机的网络故障诊断

Network Fault Diagnosis Based on Niche Particle Swarm Optimization SVM

针对支持向量机(SVM)在网络故障诊断中应用存在的参数设置和诊断模型复杂的问题,提出一种基于小生境粒子群优化的SVM解决方案。算法在进行参数寻优的同时考虑支持向量个数,实现对诊断模型复杂度的优化,并采用小生境粒子群算法进行求解,提高算法跳出局部最优的能力。在DARPA数据集上的实验表明本文提出的方法能够有效提高诊断模型的泛化性和诊断速度。

The parameters setting and diagnosis model complexity caused by dataset’s huge size affect the SVM’s application in network fault diagnosis. A new method based on niche particle swarm optimization SVM is proposed to solve these problems. The algorithm optimizes SVM training parameters,and the support vectors number is proposed to simplify the diagnosis model complexity as well. The niche particle swarm optimization method was introduced to this optimization problem while it’s excellent ability of escape locally optimal solution. The experiments on DARPA dataset shows that the method can improve the diagnosis model’s generalization and can get faster diagnosis speed.