Robust temporal low‐rank representation for traffic data recovery via fused lasso

Abstract

Achieving complete and accurate traffic data as input is crucial for most intelligent transportation systems. However, due to hardware or software malfunction, traffic data is inevitably faced with missing and noise problems. Most of the existing representationbased traffic data recovery methods adopt sparse representation theory, which well models the local association properties of traffic data, but ignores their global correlation. To overcome this shortcoming, a robust low-rank representation method that incorporates temporal prior information to impute the missing traffic data is proposed. Specifically, the low-rank representation theory is first employed to model the global spatial correlation of traffic data, and then the fused lasso regularisation is utilized to fit the temporal correlation of traffic data. In addition, to make the proposed model more robust, F-norm regularisation is used to smooth the Gaussian noise of traffic data. Furthermore, an efficient optimisation algorithm based on ADMM is developed to solve the proposed model. Finally, the extensive experiments performed on real dataset validate the effectiveness of the proposed method.

Spatiotemporal traffic data imputation by synergizing low tensor ring rank and nonlocal subspace regularization

“…Not limited to the global correlation, the local structure prior has also been widely used to characterize the strong correlations between adjacent traffic data along the spatial or temporal dimensions [36][37][38][39]. By exploiting the lower-dimensional structure embedded in traffic data, Zhou et al [40] utilized spatiotemporal within-mode regularization to capture the spatial correlation and temporal stability features.…”

Robust Fused Lasso Penalized Huber Regression with Nonasymptotic Property and Implementation Studies

“…For instance, Petersen et al [5] proposed a fused lasso additive model, in which each additive function is estimated to be piecewise constant. Mao et al [6] incorporated temporal prior information to the missing traffic data and then used the fused lasso regularization to fit the temporal correlation of traffic data. Corsaro et al [7] presented a model based on a fused lasso approach for the multiperiod portfolio selection problem.…”