Nonrecurrent traffic congestion detection with a coupled scalable Bayesian robust tensor factorization model

Nonrecurrent traffic congestion (NRTC) usually brings unexpected delays to commuters. Hence, it is critical accurately detect and recognize the NRTC in a real-time manner. The advancement of road detectors provides researchers with large-scale multivariable temporal-spatial data, which allows deep research on be conducted. However, remains challenging task construct an analytical framework through natural structural properties information can effectively represented exploited better understand NRTC. In this paper, we present novel training-free based coupled scalable Bayesian robust tensor factorization (Coupled SBRTF). couple variables including flow, speed, occupancy sharing same sparse structure. Moreover, naturally captures high-dimensional patterns data by factorization. With its entries revealing distribution magnitude NRTC, shared structure compasses sufficiently abundant about While low-rank part framework, expresses general expected conditions as auxiliary product. Experimental results real-world show that proposed method outperforms detection models principal component analysis (coupled BRPCA), rank sparsity decomposition (RSTD), standard normal deviates (SND). performs even when only weekdays are utilized, hence provide more precise estimations for daily