Predicting the compressive strength of fiber‑reinforced recycled aggregate concrete: A machine‑learning modeling with SHAP analysis

Fiber-reinforced recycled aggregate concrete (FR-RAC) has recently gained more popularity because of its advantages, high
strength, eco-friendliness, and cost-effectiveness. This study uses an advanced machine-learning technique for forecasting
the compressive strength of FR-RAC. In this study, an experimental database that contained pertinent data from several
previous research was evaluated to train and test using machine learning (ML) techniques and models. To accurately represent
the subtle interactions within the dataset, the multivariate analysis identifies and includes essential factors that impact the
complicated behavior of FR-RAC in the model. This study presents a hybrid ML model for predicting concrete’s compressive
strength by combining several machine learning algorithms in a novel way. To predict the reliability of machine learning
models, several algorithms, such as adaptive boosting regressor, support vector regressor, KNN regressor, gradient boosting,
and random forest, were developed to help find the interrelated behaviors of parameters. Among all the models used in
this study, the Light Gradient-Boosting Machine (GBM) outperforms (
R2 = 0.90) other models, each of which was fitted
to a different portion of the training dataset. Additionally, the SHAP analysis revealed that recycled coarse aggregate has
an inverse impact on the strength of FR-RAC. Overall, the outcomes of this study can significantly contribute to cost and
material reduction by predicting the compressive strength of FR-RAC without the need for extensive laboratory testing and
promoting more efficient use of resources.