Compressive Strength Prediction of Self-Compacting Concrete with Recycled Coarse Aggregate Using Machine Learning: Robust Multi-Split Evaluation and Data-Leakage Analysis of a Stacking Ensemble
Robust, leakage-aware evaluation, rather than model architecture, emerges as the key to credible strength prediction on this benchmark, with Mutual-information and partial-dependence analyses identify cement as the dominant predictor, with water acting mainly through a nonlinear dependence.
Abstract
Reliable prediction of the compressive strength of self-compacting concrete with recycled coarse aggregate (SCRCAC) from mixture composition supports more rational mix design and fewer experimental tests. Using the benchmark dataset of the reference study (603 mixtures, eight input variables), this work re-examines machine-learning prediction of this property with an emphasis on honest evaluation rather than on a new model. A stacking ensemble of three gradient-boosting models (XGBoost, LightGBM, CatBoost) and an extremely randomized trees model, combined through a ridge meta-learner, is used as a representative model and compared with the four machine-learning models of the reference study (Random Forest, Extra Trees, XGBoost, LightGBM), the recent single-booster model of Abood et al., and the reference artificial neural network. Reported as the mean over 25 repeated 70/30 splits, the ensemble reaches R2 = 0.793 ± 0.038 and RMSE = 6.21 ± 0.48 MPa, above all four reference models (R2 = 0.7249–0.7635) and significantly, though only marginally, above a tuned single XGBoost. The central contribution is the evaluation itself. Because the dataset contains repeated identical compositions, a leakage-free protocol lowers the R2 of every model to between 0.60 and 0.71, showing that the values of about 0.81–0.87 usually reported are inflated by duplicate-composition leakage, and leave-one-source-out evaluation lowers it further to about 0.14. Mutual-information and partial-dependence analyses identify cement as the dominant predictor, with water acting mainly through a nonlinear dependence. Robust, leakage-aware evaluation, rather than model architecture, emerges as the key to credible strength prediction on this benchmark.
Machine learning is increasingly used to predict the compressive strength of self-compacting concrete with recycled coarse aggregate (SCRCAC), with coefficients of determination of 0.81–0.87 reported in the literature. This paper first shows that the widely used reference dataset of 603 mixtures contains only 504 unique compositions, with 84 groups of identical and 21 contradictory ones, so that identical mixtures leak between the training and test sets. Under an objective, leakage-free evaluation (with the model re-tuned on the deduplicated dataset), the coefficient of determination drops to about 0.73, a correction that applies to all models on this dataset. We then propose an interpretable, hyperparameter-optimized LightGBM model that (i) reaches the level of the best published results under the standard protocol (seed-averaged five-fold cross-validation (CV) R2 = 0.813); (ii) provides a calibrated uncertainty interval for each prediction via split-conformal prediction, achieving an empirical coverage of 0.906 at the 90% nominal level; and (iii) remains fully explainable (SHAP (SHapley Additive exPlanations), partial dependence), with cement as the dominant predictor, followed by water and the mineral admixture. Under a 70/30 protocol averaged over 25 splits, it achieves an R2 = 0.794 ± 0.038 and a root mean squared error (RMSE) = 6.20 ± 0.48 MPa, exceeding all four machine learning models of the reference study. Aspects in which the reference study retains an advantage are also discussed.
Bojana Milošević, Nenad Kojić, Milan Kragović· Journal of Composites Scienc...· 0 citations
Abstract This study develops a robust framework for estimating the compressive strength of self-compacting concrete (SCC) incorporating recycled aggregates using supervised machine learning (ML) techniques. A comprehensive experimental database comprising 582 concrete mix designs was used, encompassing diverse input variables including binder content, water, coarse and fine aggregates, recycled aggregate proportion, superplasticizer dosage, and curing time. Seven ML algorithms—XGBoost, CatBoost, AdaBoost, Extra Trees, Bagging Regressor, K-Nearest Neighbors, and Radius Neighbors—were systematically trained using a stratified 70/15/15 data split and optimized via grid search with five-fold cross-validation. Model performance was evaluated using coefficient of determination (R 2), root mean squared error, and MAE across training, validation, and testing datasets. Among all models, XGBoost demonstrated the highest accuracy, achieving an average R 2 of 0.9799, RMSE of 2.87 MPa, and mean absolute error of 1.97 MPa. The Permutation Feature Importance analysis revealed that binder content, water, and coarse aggregate were the most influential predictors of strength. This study confirms that ensemble ML models, particularly XGBoost, can reliably predict the compressive strength of SCC with recycled aggregates, while offering transparent insights into material behavior. The results provide a valuable tool for sustainable mix design optimization and practical implementation in eco-efficient concrete construction.
A. Khan, M. D. Rasheed, Muhammad Huzaifa Naveed et al.· Data-Centric Engineering· 0 citations
Recycled aggregate concrete (RAC) is a sustainable alternative construction material to reduce natural resource exploitation and manage construction and demolition waste. However, predicting the mechanical performance of RAC remains a challenge due to the high variability of recycled aggregate properties. The purpose of this study is to develop a machine learning model to predict the compressive strength of recycled aggregate-based concrete and compare its performance with normal concrete. The dataset used consists of 2165 samples (1600 normal concrete and 565 recycled aggregate concrete) collected from various scientific publications. Three tree-based machine learning algorithms (Random Forest, XGBoost, and LightGBM) were implemented and optimized using RandomizedSearchCV with 5-fold cross-validation. The results showed that LightGBM provided the best performance with R² = 0.92, MAE = 2.45 MPa, and RMSE = 3.52 MPa on the test set. This model is able to predict the compressive strength of normal concrete (R² = 0.92) and recycled aggregate concrete (R² = 0.91) with almost the same accuracy, indicating strong generalization. Feature importance analysis revealed that curing age, cement content, and water content are the most important factors in compressive strength prediction, while for RAC, recycled aggregate water absorption (WRCA) also makes a significant contribution. Error analysis shows that residuals are random and normally distributed without systematic bias. This model can reliably predict concrete compressive strength in the range of 20-60 MPa with an average error of ±3-4 MPa and can be integrated into mix proportioning design software to improve the efficiency of the design process and support the use of sustainable construction materials.
Suji’at, Eko Wahyu Abryandoko, Ocha Silvia Kencana et al.· Journal of Novel Engineering...· 0 citations
A comparative framework evaluating nine regression algorithms using the UCI Concrete Compressive Strength dataset, jointly integrating correlation-corrected statistical validation, multi-model Bayesian optimization, and domain-informed feature engineering with SHAP interpretation, rarely combined in prior concrete-strength studies.
Musthafa 'Abduh Fakhruddin, Sri Winarno, Acun Kardianawati· IDEALIS : InDonEsiA journaL...· 0 citations
Compressive strength is the single most important design parameter governing the safety, serviceability, and economy of concrete structures, yet its determination through standard 7-, 14-, or 28-day destructive cylinder/cube testing is slow, costly, and unable to assess concrete already cast in place. This study develops and evaluates a Random Forest (RF) regression model to predict the compressive strength of concrete directly from eight standard mix-design parameters — cement, blast furnace slag, fly ash, water, superplasticizer, coarse aggregate, fine aggregate, and curing age — using Yeh's (1998) benchmark dataset of 1,030 experimentally tested concrete mixtures. Following data cleaning, exploratory correlation analysis, an 80:20 train-test split, and five-fold GridSearchCV hyperparameter tuning, the optimized Random Forest model is benchmarked against Linear Regression, Ridge Regression, and Support Vector Regression using the coefficient of determination (R²), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). The Random Forest model achieves the strongest predictive performance of the models tested, substantially outperforming the linear baselines and confirming that concrete strength development is governed by non-linear interactions among mix constituents. Feature importance analysis further shows that curing age and cement content are the dominant predictors, while water content exerts a clear negative influence consistent with Abrams' Law, and coarse/fine aggregates contribute comparatively little, consistent with their role as largely inert fillers. These findings demonstrate that Random Forest regression offers a fast, accurate, and interpretable, non-destructive alternative to conventional strength testing, with practical value for mix-design optimization, quality control, and early-stage structural decision-making.
M. Selvakumar, S. Geetha, P. K. Kumar et al.· International journal of com...· 0 citations
Self-compacting geopolymer concrete (SCGC) requires coordinated control of workability and strength during mix design, yet predicting fresh-state properties from mix proportions alone has not been reliably achieved. This study trained machine learning models on 327 two-part SCGC mixes from 37 published sources to predict five fresh properties (slump flow, T500, V-funnel time, L-box ratio, J-ring step) and 28-day compressive strength (CS₂₈d). Curing temperature, pre-demoulding duration, and post-curing regime were added as inputs for CS₂₈d, giving 15 inputs total. Five tree-based algorithms (random forest, gradient boosting, extra trees, XGBoost, LightGBM) were compared, with hyperparameters tuned via RandomizedSearchCV or Optuna. Under random-split cross-validation, extra trees achieved CV R² of 0.949 (V-funnel), 0.910 (L-box), and 0.934 (J-ring); gradient boosting led for slump flow at CV R² = 0.929. Under Leave-One-Source-Out (LOSO) validation; which withholds entire laboratories from training; R² fell to 0.261 for slump flow and 0.120 for CS₂₈d; T500 reached R² = −0.279. The resulting gap, ΔR² = 0.67–0.92 across outputs, measures the inter-laboratory information leakage that random-split validation conceals and that prior SCGC ML studies have not corrected for. Adding the curing inputs raised CS₂₈d test R² by 0.119. SHAP and permutation importance analysis identified curing temperature as the dominant driver of CS₂₈d and produced physically consistent rankings for fresh property outputs. The trained models served as surrogates in a differential evolution framework that simultaneously optimises for EFNARC workability classes (SF2/SF3, PA2) and minimum compressive strength targets, tested for ambient (25°C, 24 h) and oven (70°C, 48 h) curing across six strength thresholds. Eleven of twelve scenarios returned feasible solutions, with total binder content from 362 to 574 kg/m³.
M. Aydın, A. Kurtoglu· Engineering Research Express· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
