Traffic congestion prediction based on GPS trajectory data

Shuming Sun,Jian Sun,Juan Chen

doi:10.1177/1550147719847440

Abstract

Since speed sensors are not as widely used as GPS devices, the traffic congestion level is predicted based on processed GPS trajectory data in this article. Hidden Markov model is used to match GPS trajectory data to road network and the average speed of road sections can be estimated by adjacent GPS trajectory data. Four deep learning models including convolutional neural network, recurrent neural network, long short-term memory, and gated recurrent unit and three conventional machine learning models including autoregressive integrated moving average model, support vector regression, and ridge regression are used to perform congestion level prediction. According to the experimental results, deep learning models obtain higher accuracy in traffic congestion prediction compared with conventional machine learning models.

Highlights

With the improvement of living standards, the vehicle possession per capita continues to grow
The main ideas of this article are as follows: (1) first, the GPS trajectory data are matched to roads, and experimental road networks are selected according to map matching results; (2) the average driving speed of road networks is calculated by the results of map matching step and the calculation result is converted to time sequences and temporal–spatial image; (3) the convolutional neural network (CNN) and the recurrent neural network (RNN) series models are used to predict the average driving speed of the road networks; (4) the traffic congestion levels are classified according to the traffic congestion level standard
A map matching algorithm based on the hidden Markov model (HMM) model is used in this article to preprocess taxi GPS trajectory data

Summary

Introduction

With the improvement of living standards, the vehicle possession per capita continues to grow. A procedure shown in Figure 1 is proposed to perform traffic congestion level prediction based on GPS trajectory data. The main ideas of this article are as follows: (1) first, the GPS trajectory data are matched to roads, and experimental road networks are selected according to map matching results; (2) the average driving speed of road networks is calculated by the results of map matching step and the calculation result is converted to time sequences and temporal–spatial image; (3) the convolutional neural network (CNN) and the recurrent neural network (RNN) series models are used to predict the average driving speed of the road networks; (4) the traffic congestion levels are classified according to the traffic congestion level standard. Section ‘‘Conclusion’’ summarizes the research and points out the future research direction

Literature review

Findings

Declaration of conflicting interests