Missing Data Imputation on IoT Sensor Networks: Implications for on-site Sensor Calibration

Abstract

IoT sensors are becoming increasingly important supplement to traditional monitoring systems, particularly for in-situ based monitoring. However, data collection based on IoT sensors are often plagued with missing values usually occurring as a result of sensor faults, network failures, drifts and other operational issues. <br>

Data reconstruction applications for IoT air pollution sensor networks using graph signal processing

“…Liu et al [26] propose the use of low-rank matrix completion methods for missing imputation of air pollutants given their strong spatial correlation. Okafor et al [27] compare different machine learningbased imputation methods using the sensors' time series, as well as their impact on the posterior sensor calibration, showing the superiority of Variational Autoencoders (VAE). Mondal et al [28] develop a missing value imputation method for sensor networks based on spatio-temporal graph signal reconstruction via Sobolev smoothness.…”