高校智能电表缺失数据修复方法  

Method for repairing missing data in college smart meters

作　　者：陈庆斌 杨耿煌[1,2] 耿丽清[1,2] 苏娟[3] Chen Qingbin;Yang Genghuang;Geng Liqing;Su Juan(School of Automation and Electrical Engineering,Tianjin University of Technology and Education,Tianjin 300222,China;Tianjin Key Laboratory of Information Sensing and Intelligent Control,Tianjin University of Technology and Education,Tianjin 300222,China;College of Information and Electrical Engineering,China Agricultural University,Beijing 100083,China)

机构地区：[1]天津职业技术师范大学自动化与电气工程学院,天津300222 [2]天津职业技术师范大学天津市信息传感与智能控制重点实验室,天津300222 [3]中国农业大学信息与电气工程学院,北京100083

出　　处：《国外电子测量技术》2024年第5期136-143,共8页Foreign Electronic Measurement Technology

基　　金：国家重点研发计划(2022YFB2403002);天津市高等学校科技发展基金(2022ZD037);天津市科技计划(23YD TPJC00320)项目资助。

摘　　要：高校运行数据在采集、传输、存储过程中往往会产生数据缺失。对此,提出一种基于改进长短期记忆神经网络-链式方程多重插补法的缺失数据修复方法。采用链式方程多重插补法,通过迭代对每个缺失的属性值产生多个填补值,从而产生多个完整数据集,并进行分析优化得到一个最终的完整数据集。为提高缺失值修复精度,在长短期记忆神经网络的预测任务中,采用麻雀搜索算法进行超参数寻优,并结合均值匹配模型对缺失数据进行修复。使用北方某高校2019年数据进行验证,通过无自然缺失算例和自然缺失算例对提出方法进行评估,结果表明,在无自然缺失算例中,整体归因误差为0.106,较其他模型至少降低29.3%,验证了方法的有效性;对11.8%自然缺失率下的数据进行填补,经提出的方法填补之后的数据有效提高了高校后续运行数据的预测精度,间接验证了缺失数据填补的有效性。Data loss often occurs in the process of collection,transmission and storage of university operation data.In this regard,this paper proposes a missing data repair method based on improved long short-term memory neural network-chain equation multiple interpolation method.The chain equation multiple interpolation method is used to generate multiple filling values for each missing attribute value through iteration,thereby generating multiple complete data sets,and analyzing and optimizing to obtain a final complete data set.In order to improve the accuracy of missing value repair,in the prediction task of long short-term memory neural network,the sparrow search algorithm is used to optimize the hyperparameters,and the mean matching model is used to repair the missing data.The data of a university in the north of China in 2019 are used for verification.The method proposed in this paper is evaluated by non-natural missing examples and natural missing examples.The results show that the overall attribution error of this method is 0.106 in non-natural missing examples,which is at least 29.3%lower than other models,which verifies the effectiveness of this method.The data under the natural missing rate of 11.8%is filled.The data filled by the method proposed in this paper effectively improves the prediction accuracy of the subsequent operation data of colleges and universities,and indirectly verifies the effectiveness of missing data filling.

关 键 词：高校运行数据 缺失数据填补 链式方程多重插补 长短期记忆神经网络 

分 类 号：TN399[电子电信—物理电子学]