Electromagnetic data completion and prediction method based on tensor train

Electromagnetic data completion and prediction method based on tensor train

연구 분야: Infrastructure

학회: AISS '22: Proceedings of the 4th International Conference on Advanced Information Science and System

In residential environment, electromagnetic power density exceeding a certain value will affect people's livelihood and health. In the monitoring of electromagnetic environmental quality of residential buildings, the grid method is generally used to measure the data value of electromagnetic radiation sources, and the visualization technology is used to display the data of electromagnetic radiation sources in the region. In this paper, we use the method of randomly deploying sensor nodes to sample grid electromagnetic data, which greatly saves the deployment cost of sensor nodes. However, it will lead to data loss and pulse noise interference. Giving that the general electromagnetic data visualization diagram are local smoothing and sparse in transformation domain, we propose to use the tensor form of electromagnetic data to completion/restoration or predict the area grid that cannot be monitored based on the completion theory. The prediction model based on tensor train and algorithm are given. Experimental results show that the method can make the data smoother visually and within a certain accuracy.