Strength Evaluation of Expansive Soil Stabilized with Lead–Zinc Mine Tailings and Cement: An Artificial Intelligence Approach

Effective management and sustainable utilization of waste is germane to the growth of any nation due to the environmental protection and the economic advantages it offers. In this study, the sustainable utilization of lead–zinc mine tailings (LZMT) on a Portland limestone cement (PLC) modified expansive soil was investigated. The LZMT and PLC were blended with the expansive soil in varying proportions. Thereafter, their effect on the unconfined compressive strength (UCS) and California bearing ratio (CBR) of the expansive soil was assessed. The result obtained from the assessment revealed that the optimum combination of the LZMT and PLC that led to appreciable improvement in the UCS of the expansive soil was found when 25% LZMT and 12% PLC were added to the soil and that of the CBR was found when 20% LZMT and 12% PLC were added to the soil. Furthermore, relevant vector regression (RVR) and Gaussian process regression (GPR) were employed to develop models to predict the UCS and CBR of the improved expansive soil. Based on the statistical analysis used for the evaluation of the models, the GPR developed UCS model, whose R² values for the training and testing datasets were respectively 0.9957 and 0.9926, gave the best prediction accuracy model when compared with the UCS model developed with the RVR. In the case of the CBR, the RVR developed CBR model, whose R² values for the training and testing datasets were respectively 0.9191 and 0.8915, gave the best model when compared with the CBR model developed with the GPR.