Short-Term Traffic Flow Prediction Algorithm for Expressway Based on Long Short-Term Memory and Support Vector Regression 2020-01-5151
This paper proposed a short-term traffic flow prediction algorithm based on LSTM-SVR, which aims to use machine learning algorithms to solve the high real-time and accuracy requirements faced by short-term traffic flow prediction, and to implemented and compared multiple traffic flow prediction machine learning algorithm performance. In this method, the first 12 segments of 15-minute traffic flow data are reconstructed and normalized and input into the LSTM network. After denormalizing result from LSTM, we input it into the SVR model, and the prediction result is optimized by SVR. The final SVR model output result is the following A 15-minute traffic flow prediction value realizes real-time high-precision prediction of traffic flow. Based on the toll data of Xi’an City Ring Toll Station in Shaanxi Province from May 2018 to May 2019, after many experiments, we analyzed model performance from training speed, loss function value, prediction effect, and the value of prediction evaluation index, then determined the optimal sampling interval, time-step, loss function, and optimizer of the LSTM model. By drawing the data fitting curve, comparing and analyzing the fitting effect, using the forecast evaluation index combined with the real-time traffic flow forecast, we compared and analyzed multiple forecasting algorithms, and found that the proposed forecasting algorithm LSTM-SVR is better than LSTM, GRU, SAE, ARIMA, SVR, and the accuracy and stability have been significantly improved, R2 reached 0.982, and MAPE was 0.111. This traffic flow prediction algorithm provides strong support for traffic management personnel to judge the state of the road network, conduct traffic control and traffic flow guidance.
Citation: Guo, L., Chang, H., Cheng, X., Zhou, J. et al., "Short-Term Traffic Flow Prediction Algorithm for Expressway Based on Long Short-Term Memory and Support Vector Regression," SAE Technical Paper 2020-01-5151, 2020, https://doi.org/10.4271/2020-01-5151. Download Citation
Lan Ying Guo, Hui Chang, Xin Cheng, Jing Mei Zhou, Hong Fei Wang
Chang’an University, China
3rd International Forum on Connected Automated Vehicle Highway System through the China Highway & Transportation Society
Subscribers can view annotate, and download all of SAE's content.
Learn More »