J. Mohammadi, S.m. Taheri,
Volume 9, Issue 2 (7-2005)
Abstract
Pedotransfer functions are the predictive models of a certain soil property from other easily, routinely, or cheaply measured properties. The common approach for fitting the pedotransfer functions is the use of the conventional statistical regression method. Such an approach is heavily based on the crisp obervations and also the crisp relations among variables. In the modeling natural systems, like soil, we are dealing with imprecise observations and the vague relations among the variables. Therefore, we need an appropriate algorithm for modeling such a fuzzy structures. In the present study, the fuzzy regression approach was used in order to fit some chemical and physical pedotransfer functions. The optimum regression models with the fuzzy coefficients were obtained for modeling pedotransfer functions. Sensivity analysis was carried out by using the credibility level.
The results indicated that the fuzzy regression might be considered, as a suitable alternative or a complement to the statistical regression, whenever a relationship between variables is imprecise and generally when dealing with the errors due to a vaguness in regression models.
