A Transformed Tubal Tensor Train Decomposition Method for Internet Traffic Data Recovery and Forecast

发布时间：2026-06-03 作者：浏览次数：

Speaker:	凌晨	DateTime:	2026年6月10日（周三）上午10:00-11:00
Brief Introduction to Speaker: 凌晨，杭州电子科技大学		Place:	国交2号楼201会议室
Abstract:Recovery and forecast of network traﬃcdata from incomplete observed data is an important issue in internet engineering and management. In this talk, by fully considering the temporal stability and periodicity features in internet raﬃcdata, a novel optimization model for internet data recovery and forecast is proposed, which is based upon the newly introduced higher-order tensor decomposition form called tubal tensor train (TTT) decomposition. Moreover, by introducing auxiliary variables and penalty techniques, a relaxation of the proposed model is obtained. Then, an easy-to-operate and eﬀective algorithm for solving the relaxation model is proposed. We prove that the sequence generated by the proposed algorithm converges to a stationary point of the established relaxation model. A series of numerical experiments about the recovery of structurally missing traﬃcdata and the traﬃcdata prediction on the widely used real-world datasets demonstrate that ...