<b>Interpretable Machine Learning-Based Mathematical Models for Predicting Unconfined Compressive Strength of Lateritic Soils in Southwest, Nigeria</b>

The mining of lateritic soils is widely practiced across tropical regions, particularly in Southwestern Nigeria, where they are extensively utilized for infrastructure development. The selection of suitable lateritic soils is primarily governed by their geotechnical properties. However, determining their unconfined compressive strength (UCS), a critical parameter for geotechnical design, often requires labor-intensive, time-consuming, and costly laboratory testing. This study presents an interpretable machine learning framework for the prediction of UCS using easily measurable soil index properties, including specific gravity, liquid limit, plasticity index, linear shrinkage, and fines content. A total of 300 disturbed laterite samples were collected from 30 deposits and subjected to ASTM-standard laboratory testing. Three predictive models: artificial neural network (ANN), support vector machine (SVM), and gene expression programming (GEP) were developed and evaluated using statistical performance indices including coefficient of determination (R²), root mean squared error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and residual standard error (RSE). The proposed framework provides a reliable, cost-effective alternative to conventional UCS testing and supports sustainable geotechnical design by enabling rapid decision-making in material selection and soil stabilization planning.