A new hybrid approach to assessing soil quality using neutrosophic fuzzy-AHP and support vector machine algorithm in sub-humid ecosystem

Soil quality determination and estimation is an important issue not only for terrestrial ecosystems but also for sustainable management of soils. In this study, soil quality was determined by linear and nonlinear standard scoring function methods integrated with a neutrosophic fuzzy analytic hierarchy process in the micro catchment. In addition, soil quality values were estimated using a support vector machine (SVM) in machine learning algorithms. In order to generate spatial distribution maps of soil quality indice values, different interpolation methods were evaluated to detect the most suitable semivariogram model. While the soil quality index values obtained by the linear method were determined between 0.458–0.717, the soil quality index with the nonlinear method showed variability at the levels of 0.433–0.651. There was no statistical difference between the two methods, and they were determined to be similar. In the estimation of soil quality with SVM, the normalized root means square error (NRMSE) values obtained in the linear and nonlinear method estimation were determined as 0.057 and 0.047, respectively. The spherical model of simple kriging was determined as the interpolation method with the lowest RMSE value in the actual and predicted values of the linear method while, in the nonlinear method, the lowest error in the distribution maps was determined with exponential of the simple kriging.