Ordinary Portland cement and lime remain the default stabilizers for ground improvement that carries a substantial embodied-carbon penalty. This study evaluates a cement-free alternative: ferruginous sandy saprolites treated with polybutylene succinate (PBS, 0.2–1.0% by dry soil weight), a biodegradable aliphatic polyester, and xanthan gum (XG, 1.5%), a microbial biopolymer, cured for eight months. Because strength continues developing to 250 days, reliable prediction is essential for specification. A Multi Expression Programming (MEP) model was developed through 34 hyperparameter optimization trials and benchmarked against an already formulated Gene Expression Programming (GEP) model. Sensitivity analysis showed that larger populations and longer chromosomes improved accuracy, whereas excessive mutation rates degraded it. MEP outperformed GEP, achieving an overall R2 of 0.9612 and MAE of 49.56, while automatic feature selection discarded pH and MDD. Curing duration, PBS content, and XG dosage governed strength development. The resulting closed-form equation lets designers specify low-carbon binders without 250-day testing.