Multistep traffic speed prediction: A deep learning based approach using latent space mapping considering spatio-temporal dependencies

作者：

Highlights：

• Deep learning based approach has been developed for multistep traffic prediction.

• A neighbor selection algorithm has been used to extract spatio-temporal features.

• Latent space mapping is used to map latent features of source and target domain.

• The approach was tested on real traffic data of two cities from California, USA.

摘要

•Deep learning based approach has been developed for multistep traffic prediction.•A neighbor selection algorithm has been used to extract spatio-temporal features.•Latent space mapping is used to map latent features of source and target domain.•The approach was tested on real traffic data of two cities from California, USA.

论文关键词：Traffic speed prediction,Deep learning,Latent space mapping,Cross connected deep auto-encoders

论文评审过程：Received 23 April 2020, Revised 20 February 2021, Accepted 20 October 2021, Available online 30 October 2021, Version of Record 5 November 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.116140