Traffic Speed Prediction for Highway Operations Based on a Symbolic Regression Algorithm

• Main Authors: Li Linchao, Tomislav Fratrović, Zhang Jian, Ran Bin
• Format: Article
• Language: English
• Published: University of Zagreb, Faculty of Transport and Traffic Sciences 2017-09-01
• Series: Promet (Zagreb)
• Subjects: highway congestion; traffic state; sensor data; speed prediction; incident; symbolic regression; genetic programming
• Online Access: https://traffic.fpz.hr/index.php/PROMTT/article/view/2279

Due to the increase of congestion on highways, providing real-time information about the traffic state has become a crucial issue. Hence, it is the aim of this research to build an accurate traffic speed prediction model using symbolic regression to generate significant information for travellers. It is built based on genetic programming using Pareto front technique. With real world data from microwave sensor, the performance of the proposed model is compared with two other widely used models. The results indicate that the symbolic regression is the most accurate among these models. Especially, after an incident occurs, the performance of the proposed model is still the best which means it is robust and suitable to predict traffic state of highway under different conditions.