Application of machine learning for predicting the IRI of asphalt pavements

This study applies machine learning techniques to predict the international roughness index (IRI) of asphalt pavement using structural, performance, environmental, and traffic-related variables. Data were obtained from the long-term pavement performance (LTPP) database and Chinese pavement datasets, with 3066 asphalt pavement sections (construction number =1) selected for analysis. Model parameters were optimized using cross-validation combined with grid search. Considering the selected factors, three machine learning models, namely artificial neural networks (ANN), support vector machines (SVM), and XGBoost, were employed to predict IRI. Their performance was evaluated using R², root mean square error (RMSE) and mean absolute error (MAE). The results show that XGBoost achieved the best predictive performance (R² = 0.96, RMSE=0.08, MAE=0.05). Feature importance analysis based on XGBoost indicates that the initial IRI is the most influential factor. These results show that XGBoost can accurately predict asphalt pavement IRI and provide a reference model for pavement management systems.