基于支持向量基的网络应用层故障检测系统  

Network Application-Layer Fault Detection System Based on SVM

作　　者：李千目[1,2] 许满武[1] 张宏[3] 刘凤玉[3] 

机构地区：[1]南京大学软件新技术国家重点实验室 [2]南京理工大学计算机科学与技术系,南京210094 [3]南京理工大学计算机科学与技术系

出　　处：《系统仿真学报》2006年第7期1806-1809,共4页Journal of System Simulation

基　　金：江苏省博士后科研资助计划(0202003402);南京理工大学青年学者基金(njsut06001);南京理工大学科研发展基金(2005-2006)

摘　　要：提出基于支持向量基的网络应用层故障检测模型,并对模型各个组件的功能、机制、实现进行了深入探讨。利用异构数据集上的距离度量函数进行预处理,解决了因训练数据规模过小而导致的训练结果太差问题,提高了训练效率。在训练过程中,考虑不同的网络数据特征对检测结果的影响程度,通过加权处理,提高了检测精度。提出一种提高不同样本在训练数据中比例的方法,解决由于样本的不均衡性而导致的某一类型的检测率偏低问题,使检测率得到提高。A framework of SVM based Network Fault Detection System of Application Layer was proposed. The function, mechanism and realization of the components of this framework were discussed. By means of distance metric of heterogeneous datasets, the feature data of network were preprocessed. Based on guaranteed estimators, the size of test set was estimated. Thus the bad train result for lack of examples was not only avoided, but the training time was also reduced and the efficiency of training was improved. During the training, by means of fuzzy mathematics, considering the effect of different network data features to the classification, a weight method was brought forward. It improved the accuracy of network fault detection. The problem of low detection accuracy of some types of faults for the imbalance of training examples was researched. A method of increasing the proportion of the examples of these types was proposed. It improved the detection accuracy of these types of faults.

关 键 词：分类决策 支持向量机 网络故障诊断 专家系统 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]