What question did this study set out to answer?

The aim is to accurately predict the International Roughness Index (IRI) of asphalt pavements by integrating various factors.

April 1, 2026BuildingsOpen Access

An Interpretable Pretrained Tabular Modeling Framework for Predicting IRI Across Multiple Pavement Structural Configurations

View Full Paper

Ask AI

Bookmark

Authors

Demonstrates a new modeling approach that predicts IRI in pavement structures, suggesting improved maintenance planning.

The aim is to accurately predict the International Roughness Index (IRI) of asphalt pavements by integrating various factors.
Constructed a dataset of 10,836 samples from the Long-Term Pavement Performance (LTPP) database
Applied feature selection validated by variance inflation factor (VIF) and correlation analysis
Trained nine machine learning models, optimizing hyperparameters using Bayesian optimization
Utilized five-fold cross-validation to enhance model generalization
Developed a user-friendly graphical interface for IRI predictions
The TabPFN model achieved a coefficient of determination R2 = 0.9474 and RMSE = 0.138 on test data
TabPFN’s predictive performance is statistically superior to baseline models according to paired t-tests
Traffic load identified as the primary driver of IRI growth, while structural layer thickness can reduce roughness
Climatic factors exhibited indirect effects on IRI through environmental exposure over time
Drivers varied slightly by pavement structure, but traffic load remained dominant

Qin et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark