Predictive Modeling of Soil Electrical Resistivity Using Ensemble Machine Learning Algorithms with Geotechnical Parameters

Soil electrical resistivity testing offers a rapid, non-destructive method for geotechnical site characterization, providing key advantages over traditional intrusive techniques. While prior research has examined correlations between resistivity and individual soil properties using regression models, such approaches often fall short in capturing the complex, multivariate nature of resistivity behavior. This study evaluates the performance of three machine learning algorithms, Support Vector Regression (SVR), Random Forest (RF), and Extreme Gradient Boosting (XGB), in predicting soil electrical resistivity based on water content, plasticity index, and dry density. A total of 30 soil samples were collected from power grid substations across three regions in Thailand and analyzed under controlled laboratory conditions. Electrical resistivity was measured using the Wenner array method over a 28-day observation period. Model performance was assessed using the coefficient of determination (R ² ) and root mean squared error (RMSE). Among the three models, XGB exhibited the highest predictive accuracy (R ²  = 0.943, RMSE = 62.72 Ω·m for training; R ²  = 0.898, RMSE = 85.77 Ω·m for testing), followed by RF (R ²  = 0.655, RMSE = 154.91 Ω·m), while SVR underperformed (R ²  = 0.038, RMSE = 258.78 Ω·m). Feature importance analysis revealed water content as the most influential parameter across all models, consistent with its strong negative correlation with resistivity (Spearman’s ρ = − 0.94). Error analysis showed that model accuracy was highest for low resistivity values (< 100 Ω·m) and deteriorated significantly for high resistivity values (> 500 Ω·m), likely due to the exponential nature of the moisture–resistivity relationship. Overall, the findings demonstrate that ensemble learning models, particularly XGB, can effectively predict soil resistivity using basic geotechnical parameters, offering a cost-efficient tool for preliminary subsurface assessments in geotechnical engineering.