Low-rank tensor methods and their relaxation forms have performed excellently in tensor completion problems, including internet traffic data imputation. However, most are based on the unfolding matrix's nuclear norm, which inevitably destroys the traffic tensor structure and significantly suffers from computation burden. Also, few consider the intrinsic spatiotemporal features, especially for the underlying spatial similarity. This paper proposes a novel low-rank and spatiotemporal priors enhanced Tucker decomposition (called LSPTD) for internet traffic data imputation. LSPTD model exploits the spatial similarity using factor graph embedding and characterizes the temporal correlation using the Toeplitz matrix. Two easily implementable algorithms and the closed-form updating rules are designed to solve the LSPTD model. Numerical experiments on the Abilene and GÉANT datasets demonstrate that our proposed model is superior to the other imputation methods in terms of NMAE and computation time.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    LSPTD: Low-Rank and Spatiotemporal Priors Enhanced Tucker Decomposition for Internet Traffic Data Imputation


    Contributors:
    Gong, Wenwu (author) / Huang, Zhejun (author) / Yang, Lili (author)


    Publication date :

    2023-09-24


    Size :

    850082 byte





    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English



    Low-Rank Autoregressive Tensor Completion for Spatiotemporal Traffic Data Imputation

    Chen, Xinyu / Lei, Mengying / Saunier, Nicolas et al. | IEEE | 2022


    Low-Rank and Deep Plug-and-Play Priors for Missing Traffic Data Imputation

    Chen, Peng / Li, Fang / Wei, Deliang et al. | IEEE | 2025




    Low-Rank Tensor Completion With 3-D Spatiotemporal Transform for Traffic Data Imputation

    Shu, Hao / Wang, Hailin / Peng, Jiangjun et al. | IEEE | 2024