This study explores the application of machine learning (ML) for predicting the optimum moisture content (OMC) of soil–stabilizer mixtures. The study utilizes support vector regression (SVR), a well-established ML technique, to develop comprehensive and precise models that establish a relationship between the OMC of various properties of natural soil and stabilized soil, such as particle-size linear shrinkage, plasticity, distribution, and the kind and amount of chemicals used to stabilize. To evaluate the sensitivity of \({\text{OMC}}\) to variations in influential factors, a diverse dataset comprising different soil types and previously published stabilization testing results is employed. Additionally, the study incorporates two meta-heuristic algorithms, namely artificial rabbits optimization (ARO) and crystal structure algorithm (CSA), to improve the accuracy of the models further. These algorithms are used to validate the models by analysing OMC samples from various soil types obtained from previous stabilization test results. The findings of the study revealed three distinct models: hybrid forms of SVCS (SVR + CSA), SVAR (SVR + ARO), and an individual SVR model. Each of these models provides valuable insights that contribute to the accurate prediction of OMC for soil–stabilizer mixtures. The SVCS model demonstrates outstanding performance, evidenced by an impressive R2 value of \(0.9759\) and an exceptionally low RMSE value of \(1.184\%\) . These results not only underscore the precision and dependability of the SVCS model but also emphasize its efficacy in forecasting soil steadying results. This method introduces a promising technique for precise OMC prediction in diverse engineering applications related to soil stabilization mixtures.
