Predicting the Compressive Strength of Self-Consolidating Concrete Using Machine Learning and Conformal Inference

Fatemeh Mobasheri¹, Masoud Hosseinpoor^1,*, Ammar Yahia^1,2, Farhad Pourkamali-Anaraki³
1 Department of Civil and Building Engineering, Université de Sherbrooke, Sherbrooke, QC J1K 2R1, Canada
2 Department of Architecture, Canadian University Dubai, Dubai, P.O. Box 117781, United Arab Emirates
3 Department of Mathematical and Statistical Sciences, University of Colorado Denver, Denver, CO 80204, USA
* Corresponding Author: Masoud Hosseinpoor. Email: email

Computer Modeling in Engineering & Sciences https://doi.org/10.32604/cmes.2025.072271

Received 23 August 2025; Accepted 07 November 2025; Published online 01 December 2025

Download PDF

Abstract

Self-consolidating concrete (SCC) is an important innovation in concrete technology due to its superior properties. However, predicting its compressive strength remains challenging due to variability in its composition and uncertainties in prediction outcomes. This study combines machine learning (ML) models with conformal prediction (CP) to address these issues, offering prediction intervals that quantify uncertainty and reliability. A dataset of over 3000 samples with 17 input variables was used to train four ensemble methods, including Random Forest (RF), Gradient Boosting Regressor (GBR), Extreme gradient boosting (XGBoost), and light gradient boosting machine (LGBM), along with CP techniques, including cross-validation plus (CV+) and conformalized quantile regression (CQR) methods. Results demonstrate that LGBM and XGBoost outperform RF, improving R² by 4.5% and 5.7% and reducing Root-mean-square Error (RMSE) by 24.6% and 24.8%, respectively. While CV+ yielded narrower but constant intervals, CV+_Gamma and CQR provided adaptive intervals, highlighting trade-offs among precision, adaptability, and coverage reliability. The integration of CP offers a robust framework for uncertainty quantification in SCC strength prediction and marks a significant step forward in ML applications for concrete research.