Desenvolvimento de um algoritmo de recomendação baseado em avaliações usando grafos

Abstract

Recommendation systems are widely used in various platforms for better usability by the user. Using historical interactions of users or users with the same interests, the search is performed to present the information so that the user has a better quality when looking for information to buy, watch, listen, etc. There are several studies in the area that propose search techniques and methods that aim both at the user, making predictions, and at the items that are listed in order to be associated to each other according to their scores by the users. In the majority of the works performed, the use of collective filtering is used, which aims at presenting results according to the evaluations of user groups based on the target user, and the recommendation based on content that makes predictions based on content already evaluated or consumed by the user. This work, while analyzing this issue, proposes the creation of a recommendation with the data from a set of evaluations previously performed, thus aiming to present users with unique sets of results with recommended evaluations based on previous interactions.