Journal of Materials and Engineering Structures
Volume 12, Numéro 3, Pages 253-265
2025-09-30

Machine Learning Approaches For Predicting Flexible Asphalt Pavement Elastic Modulus: Ann Vs. Random Forest Vs. Xgboost



Authors : Chu Tien Dung . Nguyen Quang Phuc . Luong Xuan Chieu .

Abstract

This study compared four predictive models—Linear Regression (LR), Artificial Neural Networks (ANN), Random Forest (RF), and Extreme Gradient Boosting (XGBoost)—for estimating the elastic modulus of flexible asphalt pavement. ANN demonstrated the highest predictive accuracy, with an optimized architecture featuring a single hidden layer with nine neurons. It achieved the best performance, with an R² value of 0.997, RMSE of 0.677, and MAE of 0.483. LR showed the weakest results, while RF and XGBoost performed better, with XGBoost slightly outperforming RF. However, neither model surpassed ANN. The findings highlight the importance of selecting appropriate predictive models for pavement modulus estimation, as ANN effectively captured complex nonlinear relationships. Due to limited data, the study evaluated models without splitting the dataset, which may have led to an overly optimistic assessment. Future research should incorporate larger datasets from diverse pavement structures and traffic conditions to improve model generalization. Additionally, exploring hybrid modeling approaches that combine multiple machine learning algorithms could further enhance predictive accuracy, offering more robust solutions for pavement design and quality assessment.

Keywords

Machine Learning ; Elastic Modulus ; Asphalt Pavement ; Quanlity Control and Assurance

Les articles similaires

Effects Of Asphalt Binder And Aggregate Gradation On Dynamic Modulus, Resilient Modulus And Moisture Resistance Of Asphalt Concretes

Khamkhanpom Phouxay .  Nguyen Quang Tuan .  Hoang Thi Thanh Nhan .  La Van Cham .  Deuanhaksa Surya . 
pages 193-205.

Evaluation Of Subgrade Resilient Modulus And Modulus Ratio For Geotextile Reinforced Water Bound Macadam Road Using Full-scale Laboratory Accelerated Pavement Testing Set Up

Ingle Ganesh S. . 
pages 131-140.

3-d Move Mechanistic Analysis And Cost Effectiveness Of Asphalt Rubber And Polymer Modified Asphalt Pavement Under Various Axle Loading Conditions

Proffer Logan .  Souliman Mena I. . 
pages 291-298.

Random Forest Modeling For Forecasting Economic Growth: A Machine Learning Approach In The Case Of Algeria (1964-2022)

Messaoudi Malika . 
pages 391-410.

Modeling Non-linear Time Of Concentration In Heterogeneous Catchments: A Comparative Study Of Svr And Ann Machine Learning Approaches

Suljic Nedim .  Prskalo Gordan . 
pages 505-516.

Formats de citation