关键词: 公路运输, 交通流预测, 深度学习, 高速公路, 交通拥堵

Abstract:

In order to accurately predict highway traffic flow and thus alleviate traffic congestion of highway, a highway traffic flow prediction model with multi- features is proposed in this paper. Firstly, the traffic flow of the section with the upstream and downstream sections, and the weather data are transformed into a two-dimensional matrix, and the sliding window model is employed to obtain the optimal size of input samples. Then these samples are input into a hybrid depth deep learning model to extract the features of traffic flow data, and then output the prediction results. Finally, the traffic flow data of a real highway is used to do two experiments on weekdays and holidays. The results indicate that the hybrid deep learning models perform better results than single models for forecasting highway traffic flow. The prediction accuracy of highway traffic flow on weekdays is higher than that on holidays. The proposed model reduces the mean absolute error from 6.40 Cars · (20 min)- 1 to 5.450 Cars · (20 min)- 1 , which shows that the prediction accuracy of highway traffic flow can be improved by considering multiple related factors.

Key words: highway transportation, traffic flow prediction, deep learning, highway, traffic congestion