多决策树融合模型MDTF的研究  

Research on model of multiple decision trees fusion (MDTF)

作　　者：傅玥[1] 马国富[2] 许冀伟[1] 李广霞[1] 

机构地区：[1]石家庄经济学院信息工程学院,河北石家庄050031 [2]中央司法警官学院信息管理系,河北保定071000

出　　处：《计算机工程与设计》2008年第13期3391-3393,共3页Computer Engineering and Design

摘　　要：基于数据挖掘的入侵检测系统中存在着检测性能低和数据挖掘效率不高等问题。为了解决这些问题,提出了多决策树融合模型MDTF,也就是把海量数据集分成若干子数据集,在子数据集上进行挖掘形成不同的子决策树,然后用加权平均法将多棵子决策树对网络数据的检测结果进行融合形成最优判断。实验采用KDD99数据,实验结果表明,该方法可以得到较好的入侵检测性能,分布并行处理可以提高数据挖掘效率。in order to improve the detection performance and the data mining efficiency of data mining-based intrusion detection system, the method of multiple decision tree fusion （MDTF） is to divide a great large dataset into several sub-datasets, mine on sub-datasets by decision tree separately, and detect network data by different sub-decision trees, and then combine the results from multiple sub-decision trees by weighted average. Using the dataset of KDD99, the experimental results show that this technique is superior to the single decision tree of mining on a great large dataset for intrusion detection in terms of classification accuracy. Adopting distributed and parallel mining can improve the speed of data mining.

关 键 词：决策树 数据挖掘 加权平均 信息融合 入侵检测 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]