Hybrid system for video game recommendation based on implicit ratings and social networks
- PDF / 1,466,710 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 53 Downloads / 170 Views
ORIGINAL RESEARCH
Hybrid system for video game recommendation based on implicit ratings and social networks Javier Pérez‑Marcos1,2 · Lucía Martín‑Gómez1,2 · Diego M. Jiménez‑Bravo1 · Vivian F. López1 · María N. Moreno‑García1 Received: 2 December 2018 / Accepted: 3 January 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract The digital entertainment sector is one of the fastest growing in recent years. In the case of video games, the productions of some of the most popular titles are on a par with film productions. The sale of video games is in the millions, and yet there are few works on the recommendation of video games. In this work a hybrid system of video game recommendation is presented, through the use of collaborative filtering and content-based filtering, and the construction of relationship graphs. In order to improve the recommendations, a new method for estimating implicit ratings is proposed that takes into account the hours of play. The proposed recommender system improves the results of other techniques presented in the state of the art. Keywords Recommender systems · Colaborative filtering · Content-based filtering · Rating estimation · Graph-based methods · Video games
1 Introduction Many everyday tasks involve the need for making choices, e.g. what to cook, which book to read, which music to listen to or what film to watch. However, on many occasions these decisions must be made without sufficient information and personal experience, what can lead to the choice of bad alternatives. To prevent this problem, recommendations from more experienced people (or even from the Internet) are taken into account when making decisions (Resnick and * Javier Pérez‑Marcos [email protected]; [email protected]
Lucía Martín‑Gómez [email protected]; [email protected]
Diego M. Jiménez‑Bravo [email protected] Vivian F. López [email protected] María N. Moreno‑García [email protected] 1
Department of Computer Science and Automatics, Universidad de Salamanca, Plaza de la Merced, s/n, 37008 Salamanca, Spain
Department of Computer Science, Universidad Pontificia de Salamanca, C/ Compañía, 5, 37002 Salamanca, Spain
2
Varian 1997). Moreover, in many cases these recommendations entail not only a good-decision making but also the discovery of new elements. The user support process can be automated. Recommender systems (RSs) are software tools that analyze information and provide suggestions based on user interests (Ricci et al. 2015). This kind of systems aim to predict the user’s preference for a topic and the elements to be recommended are called items. Additionally, on these systems users can mark each item to express their preferences; these expressions are called ratings. There are two traditional approaches for designing RSs (Ricci et al. 2015): – Collaborative filtering The recommendation of items to the active user is based on the information that other users with similar tastes liked in the past. – Content-based recommender systems The recommendation is based on the preferences that each u
Data Loading...
