基于生成树离群检测法的用户行为提取仿真  

Simulation of User Behavior Extraction Based on Generative Tree Outlier Detection Method

作　　者：郭娜 李攀 魏荣凯[2] GUO Na;LI Pan;WEI Rong-kai(Information Engineering College,Institute of Disaster Prevention,Sanhe Hebei 065201,China;College of Computer Science and Technology,Changchun University,Changchun Jilin 130022,China)

机构地区：[1]防灾科技学院信息工程学院,河北三河065201 [2]长春大学计算机科学与技术学院,吉林长春130022

出　　处：《计算机仿真》2020年第6期257-261,共5页Computer Simulation

基　　金：河北省高等教育教学改革研究与实践项目(2016GJJG247)。

摘　　要：针对传统用户行为提取方法的用户行为数据提取结果的准确性较差、提取时间较长的问题,提出了基于生成树离群检测法的用户行为提取方法。通过数据维度信息对完整的数据集进行划分,将获取的WSPD集合进行最短连接计算,根据计算结果值构建最小生成树,从而获取K-近邻值,针对生成树检测算法提出了一种改进的离群因子算法,并通过关联、序列、聚类以及分类的方法对用户行为特征进行提取分析。仿真结果表明:生成树离群检测法能够精准提取目标信息,且提取过程简单,缩短了用户行为数据提取时间,具有提取效率快的优点。In this article,a method to extract user behavior based on spanning tree outlier detection method was proposed.The complete data set was divided by data dimension information,and then WSPD set was used to calculate the shortest connection.According to the calculation result,the minimum spanning tree was constructed,so that the K-neighbor value was obtained.Based on the spanning tree detection algorithm,an improved outlier factor algorithm was proposed.Meanwhile,the user behavior characteristics were extracted and analyzed through the methods of association,sequence,clustering and classification.Simulation results show that the spanning tree outlier detection can extract target information accurately,and the extraction process is simple.This method shortens the extraction time of user behavior data and improves the extraction efficiency.

关 键 词：生成树 离群检测法 用户行为 数据挖掘 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]