Doporučovací systém

Abstract

Nowadays, in a time of internet, users are daily overloaded by amouts of an information from various sources. Not just the amount but also their content makes effect called information overload. This thesis deals with articles recommendations for users of news portal. The work includes an overview of basic methods for recommender systems, description of processing unstructured text data, collection of users implicit ratings and recommender algorithms. A part of this work is design and implementation of hybrid recommender system and its testing on real users of news portal www.info.sk. The system consists of three main parts: user activity tracking code, server application with database storage and computational server.