基于模糊模式与决策树融合的脚本病毒检测算法  

Script Virus Detection Algorithm Based on Fusion of Fuzzy Pattern and Decision Tree

作　　者：张涛[1] 张瀚[1] 付垒朋[1] 

机构地区：[1]南开大学天津市智能机器人技术重点实验室,天津300071

出　　处：《电子与信息学报》2014年第1期108-113,共6页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61004086);高等学校博士学科点专项科研基金(200800551024)资助课题

摘　　要：构建决策树进行脚本病毒检测可以全面利用训练样本的信息,在样本特征较为复杂、样本数较大的情况下会产生大量节点,计算时间复杂度高,在剪枝过程中影响分类准确度。为融合模糊模式的信息以提高分类器性能,该文设计了决策树分类基础上的融合算法。该算法将关于模糊模式贴近度的3个特性作为决策树样本信息向量中的属性。使用训练样本集,根据上述属性在划分点上的分裂信息值及信息增益率选择分裂属性,逐步构建决策树。实验结果验证了算法的稳定性与准确度,表明这种融合方法可增加属性的区分度,减少决策树的分支数。The method using the decision tree for script virus detection can make full use of the information of training samples. But complex sample features and large number of samples will produce large number of nodes which result in the high algorithm time complexity and affect the classification accuracy due to the pruning process In order to improve classification performance, a fusion algorithm using the information of fuzzy pattern is designed based on the decision tree classification algorithm. Three important characteristics of fuzzy pattern about close degree are regarded as the three attributes of sample information vector in the decision tree to build decision tree through training get. The stability and accuracy of the algorithm is verified by experiment. The experiment results show that the proposed algorithm increases discrimination of attributes and reduces the decision tree branch.

关 键 词：脚本病毒检测 模糊模式 决策树 贴近度 

分 类 号：TP309.5[自动化与计算机技术—计算机系统结构] TP181[自动化与计算机技术—计算机科学与技术]