Generating data of an absent sensor from correlated sources

In remote healthcare, wearable biomedical sensors enable monitoring of patient activities and medical interventions remotely in real-time. However, data of some sensors or body sensor locations may be missing. Possible reasons are sensor malfunctions, transmission issues, or simply that people do not wear a particular sensor at the required body location. This may hamper the data analysis and also the ability to draw accurate medical conclusions. We investigate strategies to handle such absent sensor data. In particular, we compare K-Nearest Neighbour methods, Multivariate Imputation by Chained Equations, LightGBM Regressor, and CatBoost Regressor methods.