Multiple Regression and K-Nearest-Neighbor Based Algorithm for Estimating Missing Values within Sensor

The missing of sensor data is unavoidable when sensors are using to monitor in a system. This missing has very big effecting on the applications of sensor itself. When missing occurred, the best method is to estimate a value as closest as possible to replace it. An algorithm, named KMRA (K-Nearest-Neighbor on Multiple Regression Algorithm), builds on K nearest neighbor and multiple regression is introduced in this paper. In the process of estimation, it considers both spatial correlations from its neighbor sensor and temporal correlations from its own time serials. After computed these two correlations, the algorithm combines them into a unified estimation to be the monitored value. As algorithm KMRA counted both spatial and temporal correlations, it reaches a high efficiency and practicability in performance. The examination results show that KMRA in this paper can estimate the missing data precisely.