EXTENDED USER PROFILING APPROACHES FOR RECOMMENDATION SYSTEMS

Abstract

Currently, the volume of data in information systems and the information needs of users are increasing. This causes information overload and a number of difficulties in finding the necessary information. Therefore, individual approaches, including personalization of user profiles, are important in solving this problem. Creating a user profile is relevant for obtaining information from systems in accordance with the needs of the user and for personalizing the services provided by the system. The lack of direct user profiling in information systems creates a number of problems in providing personalized services to users. Around the world, recent research has focused on developing systems that personalize user profiles based on their data sets. The work done so far on a global scale to create and model user profiles is analyzed. In this article, mathematical algorithms such as TF-IDF, Cosine similarity, Word2Vec are used to model and personalize user data. It also provides a classification scheme for user profiling, modeling and personalization. This classification scheme is based on three components. These are user data collection, user profiling, modeling and personalization components. Additionally, the article also mentions the benefits of user identification.