Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
5770585 | Geoderma | 2017 | 8 Pages |
â¢The best influential subset on soil CEC was determined using ACO-ANFIS.â¢A subset with 5 features was select as best subset.â¢After selecting the best subset, the soil CEC was modeled using ANFIS and MLP.â¢ANFIS was more accurate than MLP in predicting soil CEC.
Soil CEC is a very important property that represents soil fertility status. Though difficult to measure, it can be predicted by soil physicochemical properties that can be easily measured. Researchers have used different input soil properties to derive pedo-transfer functions (PTFs) and predict soil CEC. To select properties that influence soil CEC, we have introduced a hybrid algorithm: an advance ant colony organization (ACO) in combination with an adaptive network-based fuzzy inference system (ANFIS). The potential power of the advance ACO-ANFIS algorithm in setting up a framework for identifying the most determinant parameters of agricultural soils CEC in an Iranian semiarid region (Rabor region, 29° 27ⲠN to 38° 54ⲠN and 56° 45ⲠE to 57° 16ⲠE) was also investigated. To make sure that ACO-AN FIS algorithm reaches its global minimum, features were selected by ANFIS. A multiple linear regression (MLR) model was constructed as benchmark for the comparison of performances. The results from ACO-ANFIS and ANFIS for feature selection and their RMSEs were the same. Results of ACO-ANFIS and ANFIS for selecting best dataset showed that five properties including soil organic matter (SOM), clay, silt, pH and bulk density (BD) had the lowest error. The ANFIS method resulted in higher model efficiency and coefficient of determination (R2 = 0.91) than MLR approach (R2 = 0.74). This study provides a strong basis for predicting soil CEC and identifying the most determinant parameters influencing soil CEC in the Iranian semiarid regions; however, its general analytical framework could be applied to other parts of the world with similar challenges.