A comparative study of LSSVR analysis on ground granulated blast-furnace slag-based concrete

Ground granulated blast furnace slag (GGBFS) can replace a part of Portland cement in concrete. The use of GGBFS can reduce CO₂ emissions that result from the production of traditional Portland cement. Compressive strength (f_c) is one of the most important factors to be considered in building and designing concrete structures since it is one of the basic requirements for concrete mixtures. The major aim of this work is to proffer an effective method for a robust assessment of machine learning techniques in the accurate prediction of concrete f_c, including the use of GGBFS. The paper tries to zero in on the application of the Least square support vector regression (LSSVR) to develop forecast models for f_c. The f_c within the databases collected ranged between 6.3 and 101.3 MPa. The RSA and JSA were the two basic search algorithms involved in the research study to develop the efficiency of the LSSVR methodologies. While many traditional methods are prone to overfitting in high-dimensional data or smaller databases, RSA and JSA differ fundamentally from older methods because they incorporate advanced, adaptive, and bio-inspired strategies that imitate natural behaviors, therefore dealing better with complex and nonlinear relationships inherent in concrete mix design. These algorithms, integrated into LSSVR, particularly address the weaknesses of previous machine learning models, such as overfitting and suboptimal convergence, by offering superior parameter tuning and better generalization of the model. Results showed that the combined LSRSA and LSJSA systems were capable of estimating. The R² values for LSRSA = 0.9963 in train, 0.9976 in validation, and 0.9969 in test, whereas LSJSA = 0.9954 in train, 0.9966 in validation, and 0.9963 in test.