A High-Dimensional Timing Data Cleaning Algorithm for Wireless Sensor Networks

Wireless Sensor Networks (WSN) use many sensor nodes to monitor various environmental information in designated areas in real-time, which has broad application prospects in many fields and industries. Due to the sensor's physical fault or technical defect, there are some errors in the collected data; therefore, it is necessary to clean and repair the data before they are used. This paper proposes a high-dimensional sequential data cleaning algorithm for WSNs. The algorithm combines the correlation between different dimensions and the temporal correlation characteristics within the same dimension. Firstly, the data is preprocessed, and the abnormal dimension is determined by combining the prior knowledge and correlation calculation. Then, the algorithm of dynamic programming and speed constraint is used to determine the outliers and mark the abnormal dimensions. Finally, the autoregressive model with exogenous variables is used to repair outliers. Experiments are carried out on a real WSN dataset in this paper. The results show that the repair effect of the proposed algorithm is better than the single dimension benchmark algorithm.