网络流量特征选择方法中的分治投票策略研究
发布时间：2018-10-07 17:24
【摘要】：特征选择作为机器学习过程中的预处理步骤,是影响分类性能的关键因素.网络流量具有数据量大,特征维度高的特点,如何快速提取特征子集,并提高分类效率对于基于机器学习的流量分类方法具有重要意义.本文提出基于分治与投票策略的特征提取方法,将数据集分裂为多个子集,分别执行特征提取算法,利用投票方法获得最后的特征子集.实验表明可有效提高特征提取的时间效率,同时使分类器取得良好的分类准确率.
[Abstract]:Feature selection, as a preprocessing step in machine learning, is a key factor affecting classification performance. Network traffic is characterized by large amount of data and high feature dimension. How to quickly extract feature subsets and improve classification efficiency is of great significance to the traffic classification method based on machine learning. In this paper, a feature extraction method based on divide-conquer and voting strategy is proposed. The data set is divided into multiple subsets, and the feature extraction algorithm is implemented separately. The last feature subset is obtained by voting method. Experiments show that it can effectively improve the time efficiency of feature extraction and make the classifier achieve good classification accuracy.
【作者单位】： 浙江大学计算机学院;浙江科技学院理学院;浙江大学图书与信息中心;嘉兴职业技术学院;
【基金】：国家973重点基础研究发展计划(No.2012CB315903) 浙江省重点科技创新团队(No.2011R50010-21,No.2013TD20) 国家自然科学基金(No.61379118) 国家科技支撑计划(No.2014BAH24F01) 国家863计划(No.2012AA01A507) 浙江省网络媒体云处理与分析工程技术中心开放课题(No.2012E10023-14)
【分类号】：TP393.06
[Abstract]:Feature selection, as a preprocessing step in machine learning, is a key factor affecting classification performance. Network traffic is characterized by large amount of data and high feature dimension. How to quickly extract feature subsets and improve classification efficiency is of great significance to the traffic classification method based on machine learning. In this paper, a feature extraction method based on divide-conquer and voting strategy is proposed. The data set is divided into multiple subsets, and the feature extraction algorithm is implemented separately. The last feature subset is obtained by voting method. Experiments show that it can effectively improve the time efficiency of feature extraction and make the classifier achieve good classification accuracy.
【作者单位】： 浙江大学计算机学院;浙江科技学院理学院;浙江大学图书与信息中心;嘉兴职业技术学院;
【基金】：国家973重点基础研究发展计划(No.2012CB315903) 浙江省重点科技创新团队(No.2011R50010-21,No.2013TD20) 国家自然科学基金(No.61379118) 国家科技支撑计划(No.2014BAH24F01) 国家863计划(No.2012AA01A507) 浙江省网络媒体云处理与分析工程技术中心开放课题(No.2012E10023-14)
【分类号】：TP393.06
【参考文献】
相关期刊论文 前1条
1 宁卓;孙知信;龚俭;张维维;;利用流量特征的GIDS报文分类优化算法[J];电子学报;2012年03期
【共引文献】
相关期刊论文 前10条
1 李斌;;网络流量分类及其现状研究[J];广西教育;2013年39期
2 周亚建;薛超;平源;;基于端口特征的P2P应用识别方案[J];北京工业大学学报;2013年11期
3 李为民;刘晓楠;缪晨;陈陆颖;雷振明;;典型业务的包长分布规律[J];电子科技大学学报;2014年02期
4 钱亚冠;张e，
本文编号：2255036
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2255036.html
