Highway traffic accident duration prediction based on ensemble learning and data augmentation

Jiaona Chen; Daofeng Li; Weijun Tao; Jing Zhang; Peng Wang

doi:10.1117/12.2671822

21 March 2023 Highway traffic accident duration prediction based on ensemble learning and data augmentation

Jiaona Chen, Daofeng Li, Weijun Tao, Jing Zhang, Peng Wang

Proceedings Volume 12609, International Conference on Computer Application and Information Security (ICCAIS 2022); 126091B (2023) https://doi.org/10.1117/12.2671822
Event: International Conference on Computer Application and Information Security (ICCAIS 2022), 2022, ONLINE, ONLINE

Abstract

Predicting the duration of traffic accidents is an important part of emergency safety management, which can provide theoretical basis for road diversion and rescue. The validity of long-term forecasts depends largely on the quality of the data. Due to the discreteness, randomness and complexity of traffic accidents, the unbalanced samples is a common problem in traditional prediction models. The generalization ability of the model is insufficient. Based on the Mixup and ensemble learning, the accident duration prediction model is designed. The new datasets are created and extended. Several machine learning techniques are compared and analyzed with RMSE, MAPE and R², including K Nearest Neighbor (KNN), decision trees (DT), random forests (RF), gradient-based trees (GBDT), AdaBoost, extreme random trees, and XGBoost. The results show that the Mixup-PSO-XGBoost performs the best as well as the training time. Furthermore, data augmentation is an effective technique to strengthen the ensemble algorithms in prediction.

Citation Download Citation

Jiaona Chen, Daofeng Li, Weijun Tao, Jing Zhang, and Peng Wang "Highway traffic accident duration prediction based on ensemble learning and data augmentation", Proc. SPIE 12609, International Conference on Computer Application and Information Security (ICCAIS 2022), 126091B (21 March 2023); https://doi.org/10.1117/12.2671822

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available