Recommender system based on the analysis of publicly available data

Abstract

A recommender system is a software system aimed to make recommendations. To be able to do that, recommender system feature several components, such as: data collection and processing, recommender model, recommendation post-processing and a user interface. Recommender systems apply one or the combination of few of the recommendation techniques. In this paper we present recommender system developed to provide users with recommendations in accordance with their interests in different domains. We deduce user interests based on his activities and posts in social network. Social network used as a source of information on user (Facebook) provides Open API allowing access to the information about the user collected on the social network. Thanks to this data we are overcoming the so-called “cold start” problem and building user profile. A recommender system is commonly associated with only one domain, while the recommender system described in this paper is able to generate recommendations from different domains (movies and music). In addition to recommendations related with the specific domain, our system is able to recommend the web articles (unstructured text), relevant to the user that may belong to more than one category of interest.