An Assessment of Tree-Based Models for Predicting Indirect Tensile Stiffness Modulus in Fiber-Enhanced Cold Mix Asphalt

This study focuses on forecasts of the Indirect Tensile Stiffness Modulus (ITSM) of fiber-reinforced Cold Mix Asphalt (CMA), using a CHAID (Chi-Squared Automatic Interaction Detector) model of tree-based machine learning methods. One of the most important advantages of the CHAID model is its simplicity when it comes to little training. This study employed 123 laboratory samples, taken on predetermined essential elements, fibre content and type, fibre length, curing time, and air voids. The CHAID model was evaluated through several performance indicators, including R² (coefficient of determination) and RMSE (Root Mean Square Error). The proposed model had an R² of 0.913, an Average Accuracy Percentage (AA%) of 16.16%, and an RMSE of 170.2 MPa. Feature importance showed both curing time and air voids as the main factors that affect ITSM. This study demonstrates the practicality of the CHAID model as a transparent and interpretable model to forecast stiffness, offering a wise choice in designing heavy-duty sustainable fiber-enhanced CMA mixtures for pavements.