广域子空间异常网络流量预测概率置信度优化

发布时间：2018-05-26 05:54

本文选题：广域子空间 + 异常网络流量　；参考：《科技通报》2015年06期

【摘要】：为了提高对网络异常流量监测模型的置信度和准确性,降低虚警概率,防止漏报。需要对异常网络流量预测概率置信度区间进行优化设计。提出一种基于预测流量数据主特征建模与广域子空间重构的异常网络流量预测算法,提高预测概率置信度。把主特征建模投影到广域子空间中,获取网络异常流量数量的对偶子梯度预测结果,其表现为一组尺度,分析网络预测流量的概率置信度的伸缩变化,选用极大似然法对网络流量的预测概率置信度进行特征建模,得到概率置信度区间。提高预测置信度和精度。实验结果表明,该算法能有效提高异常网络流量的预测精度,通过在广域子空间中特征建模,保证了预测的概率置信度,减少预测误差,性能优越。
[Abstract]:In order to improve the confidence and accuracy of the network abnormal traffic monitoring model, reduce the false alarm probability and prevent the false alarm. It is necessary to optimize the interval of probability confidence of abnormal network traffic prediction. An anomaly network traffic prediction algorithm based on main feature modeling and wide area subspace reconstruction of predictive traffic data is proposed to improve the confidence of prediction probability. The main feature modeling is projected into the wide-area subspace, and the dual sub-gradient prediction results of the number of network abnormal traffic are obtained, which are expressed as a set of scales, and the variation of probability confidence degree of network traffic prediction is analyzed. The maximum likelihood method is used to model the probabilistic confidence of network traffic prediction, and the interval of probabilistic confidence is obtained. Improve prediction confidence and accuracy. Experimental results show that the algorithm can effectively improve the prediction accuracy of abnormal network traffic. By modeling features in wide-area subspace, the probability confidence degree of prediction is guaranteed, the prediction error is reduced and the performance is superior.
【作者单位】：天津电子信息职业技术学院;
【分类号】：TP393.06

【参考文献】