Prediction of soil permeability coefficient using the GEP approach

Hydraulic permeability of soil (k) is a critical parameter for the mathematical modeling of groundwater and soil water flow. Due to the complexity of 𝑘, it is hard to develop a general empirical model which provides a reliable prediction of it. Therefore, this study used the gene expression programming (GEP) model as a powerful data-driven technique for estimating k. The available published data for estimation of 𝑘 were culled from the literature. Six effective parameters, including clay content (CC), water content (ω), liquid limit (LL), plastic limit (PL), specific density (γ), and void ratio (e), were used to establish a predictive formula to estimate k. Statistical parameters such as bias (BIAS), root mean square error (RMSE), scatter index (SI), correlation coefficient (R), and mean absolute error (MAE) were used to evaluate the accuracy of the developed GEP model. In addition, GEP findings were compared to the artificial neural network (ANN) algorithm to assess the performance of GEP. The GEP with BIAS = −0.0005, RMSE = 0.0079, SI = 57.33%, R = 0.8109 and MAE = 0.0047 outperformed the ANN with BIAS = 0.001, RMSE = 0.0090, SI = 65.12%, R = 0.7490 and MAE = 0.0053 in predicting k in the testing stage. GEP provided an explicit mathematical equation that can be utilized to determine k. Comparing the observed data and ANN results demonstrated that the GEP approach has suitable performance for predicting k.