Using Ensemble Machine Learning to Estimate International Roughness Index of Asphalt Pavements

This study utilized an ensemble machine learning algorithm to estimate the International Roughness Index (IRI) for pavement roughness evaluation. The ensemble models, including decision tree, AdaBoosting, random forest, extra tree, gradient boosting, and XGBoosting, were developed using AGE, sum ESALs, and structural number as input parameters. The random forest algorithm produced the best model with high accuracy, achieving an R2 value of 0.996 and low errors (RMSE: 0.103, MAE: 0.013, and MAPE: 4.519) on the test set. The Shapley Additive exPlanations method was employed for explainability. The findings indicate that AGE is the most influential parameter in estimating IRI. The proposed algorithm holds promise for effective pavement management system applications. End users can estimate the IRI value based on the given decisions tree for this aim.