基于特征加权的朴素贝叶斯流量分类方法研究

发布时间：2018-03-28 14:54

本文选题：流量分类(TC)　切入点：ReliefF　出处：《高技术通讯》2016年02期

【摘要】：研究了被广泛应用于互联网流量分类的朴素贝叶斯分类方法的性能特点,针对此方法在给定类别下给出的所有流量特征同等重要并且是独立的假设在现实中难以满足,致使分类准确率不高的问题,提出一种基于特征加权的朴素贝叶斯流量分类算法。该算法基于NetFlow记录的特征信息,采用特征选择算法ReliefF和相关系数方法计算每个特征的权重值,然后将网络流量分配至后验概率最大的应用类别中。实验结果表明,这种基于特征加权的朴素贝叶斯算法具有超过94%的分类准确率,并且维持了朴素贝叶斯方法简单高效、分类稳定的特性,可以满足当前高带宽网络流量分类的需求。
[Abstract]:This paper studies the performance characteristics of naive Bayes classification method, which is widely used in Internet traffic classification. In view of the assumption that all traffic characteristics given by this method under a given class are equally important and independent, it is difficult to satisfy in reality. Because the accuracy of classification is not high, a naive Bayesian traffic classification algorithm based on feature weighting is proposed. Based on the feature information recorded by NetFlow, the feature selection algorithm ReliefF and the correlation coefficient method are used to calculate the weights of each feature. Then the network traffic is allocated to the application category with the largest posterior probability. The experimental results show that the feature weighted naive Bayesian algorithm has more than 94% classification accuracy and maintains the simplicity and efficiency of the naive Bayesian method. Classification stability can meet the needs of current high bandwidth network traffic classification.
【作者单位】：中国科学院计算机网络信息中心;中国科学院大学;
【基金】：973计划(2012CB315803) 中国科学院计算机网络信息中心“一三五”计划(CNIC_PY-1401)资助项目
【分类号】：TP393.06

【相似文献】