Machine learning-based prediction of compressive strength in sustainable self-compacting concrete

The performance and durability of conventional concrete (CC) is significantly influenced by supplementary cementitious materials (SCMs) and recycled coarse aggregate (RCA). Thus, the intrusion of enrich SCMs with RCA in the cementitious matrix delivers utmost properties. This research focuses on the use SCMs and RCA in the self-compacting concrete (SCC) system and their sustainability in the development of concrete resources. Standard experimental methods for forecasting the compressive strength (CS) of SCC have constraints in terms of efficiency, time consumption, and cost. Thus, its prediction is crucial without the need for laborious experimental procedures. This work employs and evaluates a multiplicity of machine learning (ML) models to predict the CS of the cementitious matrix to tackle these issues. Thus, considered a more effective and cost-saving solution in comparison with traditional approaches. Therefore, ML approaches like, gene expression programming (GEP), decision trees (DT), and support vector regression (SVR) were employed. The performance of the model is evaluated by employing the coefficient of determination (R 2 ), statistics, and uncertainty analysis. Individual Conditional Expectation (ICE), and Partial Dependence Plot (PDP) are used to analyze the effect of parameters on strength. The findings suggest that GEP performs best achieving superior R 2 > 0.90 for the training, validation, and test data sets with a lower error. While, the uncertainty analysis shows that all modeled values lie below the threshold value. The ICE and PDP graphs confirms that cement, age, and water-cement ratio have highly relation to outcomes. The RCA replacement ratio is more important than the CA one, and SCMs play an essential role in the development of concrete compressive strength, although not as much as cement. In addition, SP depicts major contribution to SCC. Moreover, graphical user interface (GUI) is also developed to help users/researcher that will facilitate them to estimate the strength of SCC in practical applications.