基于MIC-LSTM的水体连续缺失数据插补  

Interpolation of Water Successive Missing Data based on MIC-LSTM

作　　者：周家伟 ZHOU Jiawei(Hubei Key Laboratory of Intelligent Vision Based Monitoring for Hydroelectric Engineering,China Three Gorges University,Hubei,Yichang,443002,China;College of Computer and Information Technology China Three Gorges Umvensily Hubei,Yichang,443002,China)

机构地区：[1]水电工程智能视觉监测湖北省重点实验室三峡大学,湖北宜昌443002 [2]三峡大学计算机与信息学院,湖北宜昌443002

出　　处：《长江信息通信》2023年第3期58-61,共4页Changjiang Information & Communications

摘　　要：为解决各水体水质监测站点存在大量连续数据缺失问题,提出一种MIC-LSTM预测插补组合模型:首先利用相关性分析最大信息系数(MIC)对水体监测数据中的插补目标变量与环境量之间的关系进行相关性分析,选取MIC值大于设定阈值的特征信息作为模型输入,之后引入深度学习模型长短时记忆(LSTM)神经网络挖掘输入数据特征信息,对目标变量进行预测插补。与其他插补模型进行对比实验,选用评价指标对加入相关性分析MIC方法的有效性进行评估。结果表明:加入MIC进行特征选取输入特征的LSTM模型预测准确率得到提升,具有更强的预测性能。To solve the problem of a large number of successive data missing from each water body water quality monitoring station,a MIC-LSTM predictive interpolation combination model is proposed:firstly,the correlation analysis maximum information coefficient(MIC)is used to analyze the correlation between the interpolated target variables and environmental quantities in the water body monitoring data,and the feature mfbrmation whose MIC value is greater than the set threshold is selected as the model input;after that,deep learning is introduced model Long short-term memory(LSTM)neural network is introduced to mine the input data feature information and perform predictive interpolation of target variables.By comparing the experiments with other interpolation models,evaluation metrics are selected to assess the effectiveness of the MIC method with the addition of correlation analysis.The results show that the prediction accuracy of the LSTM model with the inclusion of MIC for feature selection of input features is improved and has stronger prediction performance.

关 键 词：水体监测数据插补 时间序列 最大信息系数 长短时记忆神经网络 隔河岩水库 

分 类 号：TP398.1[自动化与计算机技术—计算机应用技术] X832[自动化与计算机技术—计算机科学与技术]