search for


A comparative study on the prediction of road surface temperature
Journal of the Korean Data & Information Science Society 2023;34:751-61
Published online September 30, 2023;
© 2023 Korean Data and Information Science Society.

Taeyong Kwon1 · Hyungchae Yoon2 · Yongho Choi3 · Sanghoo Yoon4

124Department of Statistics, Daegu University
3Department of Computer & Information Engineering, Daegu University
Correspondence to: This work was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2022R1I1A3072824).
1 Post doctor, Department of Statistics, Daegu University, Gyeongbuk 38453, Korea.
2 Undergraduate, Department of Statistics, Daegu University, Gyeongbuk 38453, Korea.
3 Assistant professor, Department of Computer & Information Engineering, Daegu University, Gyeongbuk 38453, Korea.
4 Corresponding author: Associate professor, Department of Statistics, Daegu University, Gyeongbuk 38453, Korea. E-mail:
Received July 10, 2023; Revised August 18, 2023; Accepted August 20, 2023.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License ( which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Black ice is a phenomenon that occurs when the road surface freezes thinly due to snow or rain and low atmospheric temperature. It is related to most large-scale accidents on roads in winter and causes significant damage. Therefore, predicting the road surface temperature is necessary to prevent or mitigate the impact of black ice. This study aimed to establish a machine learning model for predicting road surface temperature using moving and fixed road weather data. The machine learning models compared in this study were random forest, gradient boosting, and XGboost. The predictive performance of each model was evaluated using root mean squared error, mean absolute error, mean error, and correlation coefficient as metrics. The study data consisted of moving road meteorological data for five days (January 5, January 8, January 13, January 21, and February 5, 2020) and fixed road meteorological data observed from November 5, 2017, to December 31, 2020. A quality control algorithm was applied to the moving road meteorological data to ensure data consistency. The results showed that random forest performed the best in predicting the moving test data for the moving road surface temperature prediction model and the fixed test data for the fixed road surface temperature prediction model. However, XGboost exhibited superior predictive performance in cross-validation.
Keywords : Black ice, machine learning, quality control, road surface temperature