Multi-Constraint Time Series Imputation

Imputation of missing values in real-time univariate and multivariate series has been a critical study in recent days to produce quality data and accuracy since missing values can lead to inaccurate analysis, bias, and flawed model predictions. Monitoring sensor data logging comes with a number of challenges, most of which are brought on by potential disturbances such as power outages, communication breakdowns, or storage issues. These disruptions could lead to data gaps, which would affect the dependability of upcoming data assessments as well as the accuracy of real-time monitoring. Although several algorithms have been developed to impute the missing data, these algorithms' performance is measured with limited to one or two constraints involved in the missing value dataset. In this work, 12 imputation algorithms are analyzed with 10 datasets to determine the most effective pairings and the optimal sequence in which the algorithms will be used for a given set of predetermined constraints.