The architecture of hybrid recommender system Gefest (Generation–Expansion–Filtering–Sorting–Truncation)

Recommender system is a relatively new class of software with the task of prediction the items (movies, music, books, news, Web sites) that would be interesting to the user, through an analysis of his actions and estimates. This paper describes the construction of such system. Designed recommendation system generates a suitable user list of recommended items. The special emphasis is made on the ability to use all the available information, including the product data (description and characteristics), user data (user profile), and data about user behavior (information about the use or purchase of goods and/or explicit estimate of goods).