Novel ensemble modelling for prediction of fundamental properties of bitumen incorporating plastic waste

Plastic asphalt mixtures (PAMs) have garnered attention recently, but their field application has been limited due to a lack of understanding of asphalt mix behavior following modification. A modelling tool that can calculate the plastic influence on the characteristics of asphalt mixtures is required to close this gap. Hence, this study offers a performance analysis of various machine learning (ML) models in predicting the performance of PAMs through its various properties. These models include three methods, decision tree (DT) as an individual technique, adaboost regressor (AR), and bagging regressor (BR), as ensemble techniques for prediction of fundamental properties of PAMs i.e. air voids (Va), marshall flow (MF), marshall stability (MS), tensile strength ratio (TSR), and indirect tensile strength (ITS). A series of experimental works and their results on the PAMs properties were collected through literature, to develop ML models and compare their accuracy. The comparative findings demonstrated that the BR model had the largest coefficient of determination (R2) and the lowest statistical errors, making it the model with the best predictive ability. According to the study's findings, ensemble approaches can be effectively utilized to forecast different properties of PAMs. Comparing ensemble models to the single model that served as their base learner, ensemble models were able to increase prediction accuracy. Furthermore, cross validation was performed to check the accuracy of developed models. SHAP analysis was conducted to examine the effects of input parameters on the fundamental properties of PAMs.