DL-Traff: Survey and Benchmark of Deep Learning Models for Urban Traffic Prediction

Abstract

Nowadays, with the rapid development of IoT (Internet of Things) and CPS (Cyber-Physical Systems) technologies, big spatiotemporal data are being generated from mobile phones, car navigation systems, and traffic sensors. By leveraging state-of-the-art deep learning technologies on such data, urban traffic prediction has drawn a lot of attention in AI and Intelligent Transportation System community. The problem can be uniformly modeled with a 3D tensor (T, N, C), where T denotes the total time steps, N denotes the size of the spatial domain (i.e., mesh-grids or graph-nodes), and C denotes the channels of information. According to the specific modeling strategy, the state-of-the-art deep learning models can be divided into three categories: grid-based, graph-based, and multivariate timeseries models. In this study, we first synthetically review the deep traffic models as well as the widely used datasets, then build a standard benchmark to comprehensively evaluate their performances with the same settings and metrics. Our study named DL-Traff is implemented with two most popular deep learning frameworks, i.e., TensorFlow and PyTorch, which is already publicly available as two GitHub repositories https://github.com/deepkashiwa20/DL-Traff-Grid and https://github.com/deepkashiwa20/DL-Traff-Graph. With DL-Traff, we hope to deliver a useful resource to researchers who are interested in spatiotemporal data analysis.

Robust Design Optimization of the Cogging Torque for a PMSM Based on Manufacturing Uncertainties Analysis and Approximate Modeling

“…Compared to classical statistical methods, DNNs can better capture complex relationships in historical data while avoiding the need for hand-engineered features. Researchers have explored several approaches to represent traffic data, including temporal sequence modeling with recurrent neural networks [26], multidimensional matrix representations with convolutional neural networks [27], and graph neural networks [28]. DNN-based methods have been shown to deliver state-of-the-art results in various traffic-related tasks, such as ride-sharing [29], and travel planning [30].…”

“…In recent years, there has been a growing interest in the development of advanced predictive models for traffic information using data-driven approaches such as deep neural networks (DNNs) [5]. Multiple studies have shown that traffic forecasting models based on DNNs outperform classical machine learning methods by a large margin [6], [7]. However, these models are typically trained to minimize the averaged prediction error, resulting in a considerable variation in performance across test samples [8].…”

Graph Neural Networks in IoT: A Survey

“…Zhang et al came out with another paper [52] following the same graph design as [143], and presented a novel dynamic node-edge attention network to address the challenges from the demand generation and attraction perspectives. [39,111,112,200,301,308]. Due to the network nature of traffic flows, GNN fits the mold to model the interaction between placed sensors via GPS location or roads [114,174,288].…”

Predicting Parking Lot Availability by Graph-to-Sequence Model: A Case Study with SmartSantander

“…Deep learning on spatial and temporal data prediction. There are similar works with parking availability prediction such as predicting traffic density, pollution data, and the general availability of resources [11,18]. These works are categorized into grid-based and graph-based predictions [11].…”

“…There are similar works with parking availability prediction such as predicting traffic density, pollution data, and the general availability of resources [11,18]. These works are categorized into grid-based and graph-based predictions [11]. The grid-based prediction divides areas into equal-size grids and predicts the values for each grid (e.g., [34]).…”