A hybrid multidimensional Recommender System for radio programs. (15th July 2022)

• Record Type: Journal Article
• Authors: Fernández-García, Antonio Jesús; Rodriguez-Echeverria, Roberto; Preciado, Juan Carlos; Perianez, Jorge; Gutiérrez, Juan D.
• Is Part Of: Expert systems with applications. Volume 198(2022)
• Publication Date: 2022-07-15
• Subjects: Recommender system -- Content-based -- Ensemble of recommenders -- Hybrid recommender -- Radio programs; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2022.116706
• Languages: English
• ISSNs: 0957-4174

Abstract

Abstract: The rise of Recommender Systems has made their presence very common today in many domains. An example is the domain of radio or TV broadcasting content recommendations. The approach proposed here allows radio listeners to receive customized recommendations of radio channels they might listen to based on their specific preferences and/or historical data. Firstly, a Data Acquisition System is presented with its main task being to obtain and process data to pass to recommenders. Secondly, a dynamic hybrid Recommender System is developed based on four dimensions reflecting major aspects of radio programs: relative talk/music percentages, music genres, topics covered, and speech tone. Eight recommenders are constructed (two per dimension) using content-based or collaborative filtering algorithms depending on the nature of the data processed, whether historical data or user preferences. And thirdly, by assigning weights in accordance with the users' preferences, a dynamic ensemble of these recommenders is formed which produces the final recommendations. Experiments were carried out illustrating the usefulness of the recommendations and its acceptance by radio listeners. Highlights: Creation of a Recommender System to suggest radio programs to listeners. Make use of Pre-processed data, Users' Preferences, and Historical Data to produce recommendations. Hybrid approach using a multidimensional ensemble. Ensemble of basic recommenders that detect which recommenders use and dynamically assign weights.