{"title":"Tucker factorization-based tensor completion for robust traffic data imputation","authors":"Cheng Lyu , Qing-Long Lu , Xinhua Wu , Constantinos Antoniou","doi":"10.1016/j.trc.2024.104502","DOIUrl":null,"url":null,"abstract":"<div><p>Missing values are prevalent in spatio-temporal traffic data, undermining the quality of data-driven analysis. While prior works have demonstrated the promise of tensor completion methods for imputation, their performance remains limited for complicated composite missing patterns. This paper proposes a novel imputation framework combining tensor factorization and rank minimization, which is effective in capturing key traffic dynamics and eliminates the need for exhaustive rank tuning. The framework is further supplemented with time series decomposition to account for trends, spatio-temporal correlations, and outliers, with the intention of improving the robustness of imputation results. A Bregman ADMM algorithm is designed to solve the resulting multi-block nonconvex optimization efficiently. Experiments on four real-world traffic state datasets suggest that the proposed framework outperforms state-of-the-art imputation methods, including the context of complex missing patterns with high missing rates, while maintaining reasonable computation efficiency. Furthermore, the robustness of our model in extreme missing data scenarios, as well as under perturbation in hyperparameters, has been validated. These results also underscore the potential benefits of incorporating temporal modeling for more reliable imputation.</p></div>","PeriodicalId":54417,"journal":{"name":"Transportation Research Part C-Emerging Technologies","volume":"160 ","pages":"Article 104502"},"PeriodicalIF":7.6000,"publicationDate":"2024-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S0968090X24000238/pdfft?md5=c4c9e2411aca4cfb096d531d793cdbd7&pid=1-s2.0-S0968090X24000238-main.pdf","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Transportation Research Part C-Emerging Technologies","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0968090X24000238","RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"TRANSPORTATION SCIENCE & TECHNOLOGY","Score":null,"Total":0}
引用次数: 0
Abstract
Missing values are prevalent in spatio-temporal traffic data, undermining the quality of data-driven analysis. While prior works have demonstrated the promise of tensor completion methods for imputation, their performance remains limited for complicated composite missing patterns. This paper proposes a novel imputation framework combining tensor factorization and rank minimization, which is effective in capturing key traffic dynamics and eliminates the need for exhaustive rank tuning. The framework is further supplemented with time series decomposition to account for trends, spatio-temporal correlations, and outliers, with the intention of improving the robustness of imputation results. A Bregman ADMM algorithm is designed to solve the resulting multi-block nonconvex optimization efficiently. Experiments on four real-world traffic state datasets suggest that the proposed framework outperforms state-of-the-art imputation methods, including the context of complex missing patterns with high missing rates, while maintaining reasonable computation efficiency. Furthermore, the robustness of our model in extreme missing data scenarios, as well as under perturbation in hyperparameters, has been validated. These results also underscore the potential benefits of incorporating temporal modeling for more reliable imputation.
期刊介绍:
Transportation Research: Part C (TR_C) is dedicated to showcasing high-quality, scholarly research that delves into the development, applications, and implications of transportation systems and emerging technologies. Our focus lies not solely on individual technologies, but rather on their broader implications for the planning, design, operation, control, maintenance, and rehabilitation of transportation systems, services, and components. In essence, the intellectual core of the journal revolves around the transportation aspect rather than the technology itself. We actively encourage the integration of quantitative methods from diverse fields such as operations research, control systems, complex networks, computer science, and artificial intelligence. Join us in exploring the intersection of transportation systems and emerging technologies to drive innovation and progress in the field.