Missing Data Completion for Network Traffic with Continuous Mutation Based on Tensor Ring Decomposition

The completion of missing network traffic is of great significance for network operation and maintenance. In recent years, the low-rank tensor completion (LRTC) techniques based on tensor ring (TR) decomposition have attracted much attention. In general, the LRTC model requires the stability of the whole tensor space. However, continuous mutation of network traffic is very common in real networks. At this time, existing completion work has difficulty in capturing the global low-rank feature of normal data and ignores the local continuous feature of mutation data, leading to a decrease in completion performance. To solve the above problems, we propose a low-rank tensor completion model that can adapt to various continuous mutation patterns of network traffic. The original tensor is represented as the sum of a normal tensor and a mutation tensor to extract their features respectively. Then, an algorithm based on the alternating direction method of multipliers (ADMM) is developed to solve the proposed model. Finally, our experimental results on both synthetic and real datasets indicate that our model can adapt to various missing data completion under different continuous mutation patterns, and has more accurate completion performance compared to advanced models.