Gradient boosting approach to predict complex modulus of GO-modified asphalt at low and medium temperature

Graphene oxide (GO)-modified asphalt presents a potential solution for enhancing the stiffness and resistance to shear deformation in asphalt materials. Methods currently employed to determine the complex modulus (G*) of GO-modified asphalt, critical for performance grade (PG) classification, are noted for their rigor, expense, and time-consuming nature. In this study, a novel approach employing classical gradient boosting regression (CGB) machine learning (ML) modeling is proposed for predicting G* at low and medium temperatures. Utilizing a database comprising 201 experimental results with nine input parameters, the finely tuned CGB model achieves a coefficient of determination R² of 0.993, reflecting an exceptional predictive accuracy and its robustness in estimating G* for GO-modified asphalt. Comparative analysis with experimental data and four alternative ML algorithms highlights the superior performance of the model. Additionally, Partial Dependence Plots analysis is utilized to elucidate correlations between key input variables and G*. Lastly, a graphical user interface is constructed and further enhances the accessibility and usability of the developed predictive tool. This research not only advances predictive modeling in asphalt engineering but also provides useful insights into the influential factors governing the mechanical properties of GO-modified asphalt.