Tensor Completion Using High-Order Spatial Delay Embedding for IoT Multi-Attribute Data Reconstruction

Restricted by various factors, the data collected by sensor nodes in some Internet of Things (IoT) can only provide spatio-temporal low-resolution multi-attribute information of the monitored area. Estimating environmental data in sensorless deployment locations to achieve spatio-temporal high-resolution multi-attribute data sensing has become an urgent problem. Existing IoT data reconstruction methods either suffer from performance degradation due to continuous data loss or ignore the correlation among multi-attribute data. To overcome these two shortcomings, a multi-attribute data reconstruction method utilizing a high-order spatial delay-embedding transform is proposed in this work. Strict low-rank property can be achieved in the proposed method without additional constraints, avoiding overcomplicating the model by combining too many constraints. The tensor ring decomposition is used to approximate the rank of the formulated data and to efficiently solve the tensor completion model via the alternating least squares algorithm. Experimental results on IoT data demonstrate that the proposed method outperforms the state-of-the-art low-rank-based methods on multi-attribute data reconstruction.