(Springer International Publishing, 2014) Lyubchyk, Leonid; Grinberg, Galina
The problem of expert preference function reconstruction in decision making process of multicriterion comparative assessment of set of object is considered. The problem is reduced to integral indicator identification using available data of object’s performance indexes measurements as well as expert estimation of integral indicators values for each object and feature weights. Based on machine learning approach and expert estimations concordance technique, the solution of preference function recovering problem is obtained in the form of optimal nonlinear object feature convolution.
The proposal generalization of expert estimates concordance idea for the case of nonlinear preferance function guaranties on optimal concordance of mesuarement and expert data, whereas machine learning approach ensure the possibility of more accurate approximation expert preference function with complex structure.
