Transactions in GIS. 2020;00:1–20. wileyonlinelibrary.com/journal/tgis | 1 © 2020 John Wiley & Sons Ltd 1 | INTRODUCTION Traffic forecasting is concerned with estimating future traffic conditions (such as the density of vehicles and their speed) to enable the prediction of future events (such as congestion or travel duration) by analyzing historical traffic conditions and patterns. Highly accurate forecasts provide guidance to decision-makers, provide safety and convenience for citizens, and reduce environmental impacts. DOI: 10.1111/tgis.12644 RESEARCH ARTICLE Traffic transformer: Capturing the continuity and periodicity of time series for traffic forecasting Ling Cai 1 | Krzysztof Janowicz 1 | Gengchen Mai 1 | Bo Yan 2 | Rui Zhu 1 1 Department of Geography, University of California, Santa Barbara, CA, USA 2 LinkedIn, Mountain View, CA, USA Correspondence Ling Cai, Department of Geography, University of California, Santa Barbara, CA 93106-9010, USA. Email: [email protected] Abstract Traffic forecasting is a challenging problem due to the com- plexity of jointly modeling spatio-temporal dependencies at different scales. Recently, several hybrid deep learning models have been developed to capture such dependen- cies. These approaches typically utilize convolutional neural networks or graph neural networks (GNNs) to model spa- tial dependency and leverage recurrent neural networks (RNNs) to learn temporal dependency. However, RNNs are only able to capture sequential information in the time series, while being incapable of modeling their periodicity (e.g., weekly patterns). Moreover, RNNs are difficult to par- allelize, making training and prediction less efficient. In this work we propose a novel deep learning architecture called Traffic Transformer to capture the continuity and periodicity of time series and to model spatial dependency. Our work takes inspiration from Google’s Transformer framework for machine translation. We conduct extensive experiments on two real-world traffic data sets, and the results dem- onstrate that our model outperforms baseline models by a substantial margin.
Traffic transformer: Capturing the continuity and periodicity of time series for traffic forecasting
Jul 04, 2023
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Related Documents
