Novel prediction of unconfined compressive strength with least square support vector regression coupled with meta-heuristic optimizers

This article presents a unique approach employing Least Square Support Vector Regression ( LSSVR ) analysis to anticipate the soil-stabilizer compounds’ Unconfined Compressive Strength ( UCS ). The study shows a strong relationship between UCS and important intrinsic soil characteristics, such as linear shrinkage, particle size distribution, and the kind and amount of stabilizing additives. Custom LSSVR prediction models take a thorough approach to address these factors. Careful considerations are put into creating these models to guarantee accurate UCS estimation. This study combines 2 meta-heuristic algorithms, Sand Cat Swarm Optimization (SCSO) and Coronavirus Herd Immunity Optimizer (CHIO), to increase accuracy. These algorithms examine UCS samples from different kinds of soil and contrast the model's predictions with previous stabilization experiment results. 3 distinct models are produced by the study: the LSSC, LSCH, and an independent LSSV . All of these models provide valuable insights that aid in precise UCS forecasts. Notably, the LSSC model outperforms the others with remarkable statistical measures such as an exceptionally low RMSE value of 90.72 and an amazing R 2 value of 0.995. These results demonstrate the LSSC model's precision, durability, and predictive capacity. This strategy demonstrates a workable method for precise UCS forecasting in a range of engineering applications involving soil-stabilizer blends and highlights the significant improvement attained by adding meta-heuristic algorithms. This improvement results in more accurate forecasts, significantly impacting the construction sector.