多用户源头无线传感网络不完整数据挖掘算法  

Incomplete Data Mining Algorithm for Multi-User Source Wireless Sensor Networks

作　　者：左丽娜[1] 刘小贞 李伟杰[1] 何首武[3] ZUO Lina;LIU Xiaozhen;LI Weijie;HE Shouwu(Department of Software and Big Data,Handan Polytechnic College,Handan Hebei 056001,China;Department of Computer Science,Handan Polytechnic College,Handan Hebei 056001,China;Department of Computer Application,Guilin University of Technology,Nanning Guangxi 530001,China)

机构地区：[1]邯郸职业技术学院软件与大数据系,河北邯郸056001 [2]邯郸职业技术学院计算机系,河北邯郸056001 [3]桂林理工大学计算机应用系,广西南宁530001

出　　处：《传感技术学报》2024年第8期1454-1459,共6页Chinese Journal of Sensors and Actuators

基　　金：2019年度广西高校中青年教师基础能力提升项目(2019KY0270)。

摘　　要：针对无线传感网络多用户源头数据受噪声影响,导致数据缺失的问题,为了提高数据的完整性,提出多用户源头无线传感网络不完整数据挖掘算法。采用组合广义形态滤波方法对多用户源头无线传感网络数据展开去噪处理,避免噪声数据影响数据填补结果;采用集成学习方法对数据进行深度挖掘,将挖掘出的数据展开分类处理;利用低秩矩阵填充理论对分类后的数据展开首次填补,在此基础上引入曲线相似分类对缺失数据进行二次填补,完成多用户源头无线传感网络数据的完整挖掘。仿真结果表明,所提方法在不同数据集中获得的均方根误差低于0.164%,信噪比高于41.8dB,补全后的数据平均绝对误差为0.023%、平均百分比误差为3.5%、均方根误差为0.021%。因此,所提方法具有较好的去噪效果和较高的数据填补性能。To improve the integrity of data,an incomplete data mining algorithm for multi-user source wireless sensor networks is pro-posed.The combined generalized morphological filtering method is used to denoise the multi-user source wireless sensor network data to avoid the noise data affecting the data filling results.The integrated learning method is used to deeply mine the data,and the mined data is classified and processed.The low rank matrix filling theory is used to fill the classified data for the first time.On this basis,curve similar classification is introduced to fill the missing data for the second time to accomplish the complete mining of multi-user source wireless sensor network data.The simulation results show that the root mean square error obtained by the proposed method in different data sets is lower than 0.164%,the signal-to-noise ratio is higher than 41.8dB,the average absolute error of the data after completion is 0.023%,the average percentage error is 3.5%,and the root mean square error is 0.021%.Therefore,the proposed method has better de-noising effect and higher data filling performance.

关 键 词：无线传感网络 组合广义形态滤波方法 集成学习 曲线相似分类 数据挖掘 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]