A temporal and spatial correlation based missing values imputation algorithm in wireless sensor networks

In wireless sensor networks, the missing of the sensing data is inevitable due to the inherent characteristic of wireless sensor networks, and it causes many difficulties in various applications. To solve the problem, the best way is to estimate the missing values as accurately as possible. In this paper, a temporal correlation based missing values imputation algorithm is proposed firstly. It adopts linear interpolation model to estimate the missing values and a good estimation effect can be achieved for the sensing data changing smoothly in a short time. Next, a spatial correlation based missing values imputation algorithm is proposed. It adopts multiple regression model and estimates the missing values with the data of multiple neighbor nodes jointly rather than independently, so that it can achieve a good estimation effect even for the sensing data that changing non-smoothly. Besides, it can not only give the estimated values of the missing data, but also give the confidence interval of each missing data for the given confidence level. Based on these two algorithms, an adaptive temporal and spatial correlation based missing values imputation algorithm is proposed at the end of this paper. It performs well both for the sensing data changing smoothly and non-smoothly. Experimental results on a real-world dataset show that the proposed algorithms can estimate the missing values accurately.