Forecasting the self-healing capacity of engineered cementitious composites using bagging regressor and stacking regressor

Engineered cementitious composite (ECC) is a unique product that can significantly contribute to self-healing when continuously hydrated. To measure its self-healing capacity, it is necessary to evaluate the crack-width once the healing process is complete. ECC has a remarkable ability to self-heal, but predicting its self-healing potential is challenging. In this study, two different ensemble machine learning (ML) algorithms i.e. bagging regressor (BR) and stacking regressor (SR) were employed to estimate ECC's self-healing capacity. Model effectiveness was assessed using error analysis and k-fold cross-validation methods. The SR model had a higher R² and was more successful in predicting the outcomes than the BR model. However, both ensemble models with smaller error values also showed improved model performance. Furthermore, the crack-healing characteristics of wheat straw ash, rice husk ash, and pumice powder are recommended to be evaluated in future studies using ML methods.