Abstract:
In this paper, we propose relevance and degree of relevance of user's or item's contents. Heuristic motivations of traditional recommender systems assume that there is transitivity property of relevance. We introduce importance of transitivity property, its influence on values of evaluation metrics and show that transitivity not performed in general case for traditional recommender systems. (In Russian).
Key words and phrases:traditional recommender system, similarity measure, transitivity, content's relevance, evaluation metric.