光纤网络异常数据特征分类研究

Feature classification of abnormal data in optical fiber network

对光纤网络异常数据特征进行分类,可有效确定光纤网络出现异常的原因,提高光纤网络安全,降低其受到网络入侵的风险。目前对光纤网络中异常数据特征分类的方法未考虑异常数据的分布特点及复杂性,易陷入局部最优解,使得异常数据特征分类效果较差,不利于异常数据检测。提出基于支持向量机的光纤网络中异常数据特征分类研究。选取数据不同异常特征,以异常数据特征选择概率作为决策变量,利用粒子群方法对决策变量进行优化求解,产生微粒的引导,确定数据集分布性,跳出局部的最优,利用重采样、数据特征空间等一些相关的理论,经过数据分层抽样,利用支持向量机分类分布性光纤网络中数据异常特征。实验的结果表明,利用所提方法分类数据异常特征,能有效提高数据异常特征分类的效率与性能,为后续数据异常检测提供依据。

Classifying the abnormal data features of optical fiber network can effectively determine the reason of the fiber network anomaly,improve the security of optical fiber network and reduce the risk of network intrusion.At present,the method of abnormal data feature classification in optical fiber network does not consider the distribution and complexity of abnormal data,so it is easy to fall into the local optimal solution,which makes the classification effect of abnormal data feature worse,which is not conducive to abnormal data detection.In this paper,the feature classification of abnormal data in optical fiber network based on support vector machine(SVM)is proposed.Different abnormal features of data are selected,and the probability of feature selection of abnormal data is used as decision variable.Particle swarm optimization method is used to optimize the solution of decision variables.Generate particle guidance,determine the distribution of data sets,fix disturbed particles,jump out of local optimality,use resampling,data feature space and other relevant theories,through data stratified sampling,Support vector machine(SVM)is used to classify data anomalies in distributed fiber networks.The experimental results show that the proposed method can effectively improve the efficiency and performance of data anomaly feature classification and provide the basis for the subsequent data anomaly detection.