Truncated tensor Schatten <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="d1e3777" altimg="si99.svg"><mml:mi>p</mml:mi></mml:math>-norm based approach for spatiotemporal traffic data imputation with complicated missing patterns

Rapid advances in sensor, wireless communication, cloud computing and data science have brought unprecedented amount of to assist transportation engineers researchers making better decisions. However, traffic reality often has corrupted or incomplete values due detector communication malfunctions. Data imputation is thus required ensure the effectiveness downstream data-driven applications. To this end, numerous tensor-based methods treating problem as low-rank tensor completion (LRTC) been attempted previous works. tackle rank minimization, which at core LRTC, most aforementioned utilize nuclear norm (NN) a convex surrogate for minimization. over-relaxation issue NN refrains it from desirable performance practice. In paper, we define an innovative nonconvex truncated Schatten p-norm tensors (TSpN) approximate impute missing spatiotemporal under LRTC framework. We model into third-order structure (time intervals,locations (sensors),days) introduce four complicated patterns, including random three fiber-like cases according mode-n fibers. Despite nonconvexity objective function our model, derive global optimal solutions by integrating alternating direction method multipliers (ADMM) with generalized soft-thresholding (GST). addition, design truncation rate decay strategy deal varying scenarios. Comprehensive experiments are finally conducted using real-world datasets, demonstrate that proposed LRTC-TSpN performs well various cases, meanwhile outperforming other SOTA models almost all