多维稀疏数据流异常数据关联挖掘仿真  

Multidimensional Sparse Data Flow Anomaly Data Association Mining Simulation

作　　者：陈少波 CHEN Shao-bo(School of Mathematics and Computer Science Shanxi University of Technology,Hanzhong Shanxi 723000,China)

机构地区：[1]陕西理工大学数学与计算机科学学院

出　　处：《计算机仿真》2019年第9期342-345,共4页Computer Simulation

基　　金：基于成对约束的半监聚类并行化算法研究(SLGKY2017-08)

摘　　要：针对当前方法异常数据集数量占用内存大、异常数据关联挖掘运行时间长,导致异常数据关联挖掘效率低的问题,提出多维稀疏数据流异常数据关联挖掘方法。首先对多维稀疏数据流异常数据进行降维处理,通过最大间隔准则查询最佳鉴别向量,求得最小最大概率以分离变量的超平面。利用频域谐振幅度等条件构建异常数据的频域模型,获取数据的最大隶属度,建立异常数据信息特征模型;在此基础上,计算其影响函数,根据计算结果对数据的异常值展开分析,采用非线性函数对其投影变换,通过相关联拉格朗日乘数的值来完成对异常数据的关联挖掘。实验结果表明,提出方法在对异常数据关联挖掘时,异常数据集数据量占用内存容量较小,并且异常数据关联挖掘的运行时间短。At present,the efficiency of abnormal data association mining is low.Therefore,an association mining method of multidimensional sparse data stream abnormal data was proposed.Firstly,the dimension of abnormal data in multidimensional sparse data stream was reduced,and the optimal discriminat vector was queried by maximum interval criterion.Then,the min-max probability was obtained to separate the hyperplane of variable.After that,the frequency-domain resonance amplitude and other conditions were used to build the frequency domain model of abnormal data and the maximum membership degree of data was obtained.Moreover,the model of abnormal data information feature was established.On this basis,the influence function was calculated.According to the calculation results,the abnormal value of data was analyzed.The non-linear function was used to perform the projection transformation on abnormal data.Finally,the values of correlated Lagrange multipliers were used to complete the association mining for abnormal data.Simulation results show that,when the proposed method is applied to the association mining of abnormal data,the abnormal data sets occupies less memory.Meanwhile,the running time of association mining of abnormal data is shorter.

关 键 词：多维稀疏数据流 异常数据 关联挖掘 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]