Doklady Belorusskogo gosudarstvennogo universiteta informatiki i radioèlektroniki (Jun 2019)
A FORECASTING MODEL ON THE BASIS OF A FUZZY LEARNING SET
Abstract
A problem of constructing a numeric forecasting evaluator on the basis of a fuzzy learning set is considered. The stated general problem is connected to the definition of the missing fuzzy vector co-ordinates and their evaluation. The general formulation is divided into two tasks: to build a method producing missing fuzzy forecasting values with expected value of a fuzzy measure and forecasting quality estimation. The given mathematical backgrounds are based on the model of a multidimensional crisp classifier and its usage for the fuzzy measure definition with the following evaluation on the basis of the fuzzy vectors probabilities by R. Yager.