Short-term traffic flow prediction based on 1DCNN-LSTM neural network structure

• DOI: 10.1142/s0217984921500421
• Publication Date: 2020-10-09T15:19:03Z
• Authors: Yihuan Qiao; Ya Wang; Changxi Ma; Ju Yang
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

In the past decade, number of cars in China has significantly raised, but traffic jam spree problem brought great inconvenience to people’s travel. Accurate and efficient flow prediction, as core Intelligent Traffic System (ITS), can effectively solve problems travel management. The existing short-term prediction researches mainly use shallow model method, so they cannot fully reflect characteristics. Therefore, this paper proposed a method based on one-dimensional convolution neural network long memory (1DCNN-LSTM). spatial information data is obtained by 1DCNN, then time LSTM. After that, space-time features are used regression predictions, which input into Fully-Connected Layer. end, corresponding results current calculated. past, most survey or virtual data, lacking authenticity. paper, real will be for research. provided OpenITS open platform. Finally, compared with other road forecasting models. show that structure 1DCNN-LSTM further improve accuracy.