Abstract:
A model and method of generating context-driven recommendations for recommendation systems with multi-criteria ratings are proposed, applicable when the user's attitude to the object is fixed not by using one integral criterion (assessment, overall rating), but by using a set of individual criteria that evaluate different aspects of the object. The proposed model and method allow to solve two main problems of using recommender systems: to rank objects according to the predicted subjective integral utility with given weights of partial criteria and to rank objects according to the predicted subjective integral utility in a given context.