Deep learning is a learning algorithm which can simulate the human brain’s multi-layer perception structure to extract the deep characteristics of the data. Therefore, deep learning can extract the deep laws of traffic congestion from a large number of continuously updated traffic detection data, which can effectively compensate for the lack of long-term efficacy and expansion ability of the existing traffic congestion prediction model. As one of deep learning methods, the convolutional neural network (CNN) has the advantages of shorter forecast time and less weight parameters needed to train. This paper applies the CNN to solve the highway traffic congestion warning problem. Based on the existing research, the paper extracts the impact factors of the traffic congestion, such as traffic flow, weather, and light, and constructs the state matrix to express the state of the traffic flow. A CNN prediction model of traffic congestion is proposed in this paper, which uses the state matrix as input variables. The accuracy of the model is validated by the test samples, and the results indicate that the CNN model is more effective than the traditional neural network model to predict traffic congestion.
Prediction Model for Traffic Congestion Based on the Deep Learning of Convolutional Neural Network
17th COTA International Conference of Transportation Professionals ; 2017 ; Shanghai, China
CICTP 2017 ; 2494-2505
2018-01-18
Conference paper
Electronic Resource
English
Convolutional neural network for recognizing highway traffic congestion
Taylor & Francis Verlag | 2020
|A Hybrid Deep Convolutional Neural Network Approach for Predicting the Traffic Congestion Index
DOAJ | 2021
|