<p>This article outlines a new approach to predict the Unconfined compressive strength (UCS) of soil stabilization blends more accurately. It applies the Decision tree (DT) algorithm to develop accurate and comprehensive models. The developed model here utilizes several natural soil parameters, such as type and dosage of stabilizer additives, plasticity, linear shrinkage, and particle distribution. It can be predicted that the DT algorithm will provide better forecasting accuracy by developing meaningful correlations between these properties and UCS in stabilized soil. Two new, efficient meta-heuristic algorithms, the Tunicate swarm algorithm (TSA) and the Sea-horse optimizer (SHO), have been integrated into the analysis to enhance the precision of the model. These are the DTSH, DTTS, and DT hybrid models. The DTSH model outperforms the other models developed for the study, with superior predictive abilities and exceptional generalization following an intensive evaluation based on a wide range of soil types compiled from previous stabilization test results. The DTSH model boasts remarkable R² values of 0.996 and a perfect RMSE of 67.89 during training, indicating exceptionally high accuracy and reliability. In general, these strategies hold great promise for the precise prediction of UCS in soil stabilization blends for a wide variety of engineering applications. The models also become more accurate when combined with metaheuristic algorithms, hence making more reliable predictions. Such developments significantly impact the building sector, where soil stabilization is essential for building resilient and long-lasting infrastructure.</p>