Predicting Marshall stability and flow parameters in asphalt pavements using explainable machine-learning models

The traditional method for determining the Marshall stability (MS) and Marshall flow (MF) of asphalt pavements is laborious, time consuming, and costly. This study aims to predict these parameters using explainable machine-learning techniques. A comprehensive database comprising 721 hot mix asphalt (HMA) data points was established, including variables such as aggregate percentage, asphalt content, and specific gravity. Models were constructed using the PyCaret Python library, and their performance was assessed using metrics such as the mean absolute error (MAE) and coefficient of determination (R²). The CatBoost regression model outperformed the other models, achieving R² values of 0.835 and 0.845 for MS and MF, respectively. Additionally, Shapley values were used to quantify the variable effects on the predictions. This approach enables the efficient preselection of design variables, reducing the need for extensive laboratory testing and promoting sustainable construction practices.