prediction of soil cation exchange capacity using support vector regression optimized by genetic algorithm and adaptive network-based fuzzy inference system

Soil cation exchange capacity (cec) is a parameter that represents soil fertility. being difficult to measure, pedotransfer functions (ptfs) can be routinely applied for prediction of cec by soil physicochemical properties that can be easily measured. this study developed the support vector regression (svr) combined with genetic algorithm (ga) together with the adaptive network-based fuzzy inference system (anfis) to predict soil cec based on 104 soil samples collected from soil surface under four different land uses. the database was randomly split into training and testing datasets in proportion of 70:30. the results showed that both models were accurate in predicting the soil cec; however, comparison of the performance criteria indicated that svr results (r^2=0.84, rmse=3.21 and mape=7.62) was more accurate than anfis results (r^2=0.81, rmse=3.38 and mape=10.31). the results of sensitivity analysis showed that two parameters had the highest effect on both models were soil organic matter and clay content.