多尺度分类挖掘算法  

Multi-scale classification algorithm

作　　者：张璐璐 赵书良[1,2,3] 田真真 陈润资 Zhang Lulu;Zhao Shuliang;Tian Zhenzhen;Chen Runzi(College of Computer&Cyber Security,Hebei Normal University,Shijiazhuang 050024,China;Hebei Provincial Engineering Research Center for Supply Chain Big Data Analytics&Data Security,Hebei Normal University,Shijiazhuang 050024,China;Key Laboratory of Network&Information Security,Hebei Normal University,Shijiazhuang 050024,China;School of Mathematical Sciences,Hebei Normal University,Shijiazhuang 050024,China)

机构地区：[1]河北师范大学计算机与网络空间安全学院,石家庄050024 [2]河北师范大学河北省供应链大数据分析与数据安全工程研究中心,石家庄050024 [3]河北师范大学河北省网络与信息安全重点实验室,石家庄050024 [4]河北师范大学数学科学学院,石家庄050024

出　　处：《计算机应用研究》2021年第2期414-420,共7页Application Research of Computers

基　　金：国家社科基金重大项目(13&ZD091,18ZDA200)。

摘　　要：多尺度分类挖掘多局限于空间数据,且对一般数据尺度特性进行分类的研究较少。针对上述问题,进行普适的多尺度分类方法研究,以扩大多尺度适用范围。从空间数据估计角度出发,结合层次理论和尺度特性,基于概率密度估计离散化方法,针对数据的多尺度特性进行分类挖掘。以非局部均值和三次卷积插值为理论基础,利用Q统计和不一致度量进行操作,提出多尺度分类尺度上推算法和多尺度分类尺度下推算法。采用UCI数据集和H省人口真实数据集进行实验,并与CFW、MSCSUA和MSCSDA等算法进行对比,结果表明,该算法可行有效。与其他算法相比,尺度上推算法正确率平均提高4.5%,F-score提高4.8%,NMI提高12.3%,尺度下推算法各个相应指标分别平均提高5.3%,6.6%和11.8%。Multi-scale classification mining are mostly limited to spatial data,and there are few researches on scale characte-ristics of general data.By solving the above problems,this paper tried to study the universal multi-scale classification method,in order to expand the scope of multi-scale application.From the perspective of spatial data estimation,combined the hierarchical theory and scale characteristics,and based on the discretization method of probability density estimation,this paper studied the classification mining on multi-scale characteristics of general data.Based on the theory of non-local mean and double cube interpolation,using Q statistics and inconsistent measurement to operate,it proposed the upscaling algorithm of multi-scale classification and downscaling algorithm of multi-scale classification.This paper performed experiments on UCI data sets and H province real population data set,and compared with CFW,MSCSUA,MSCSDA and other algorithms.The results show that the algorithms in this paper are feasible and effective.Compared with other algorithms,the upscaling algorithm improves accuracy by 4.5%,F-score by 4.8%and NMI by 12.3%and the downscaling algorithm improves the correspon-ding indexes by 5.3%,6.6%and 11.8%.

关 键 词：多尺度 不一致度量 尺度转换 多尺度分类挖掘 Q统计 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]