Recommender System for Free-to-play Games

The online platforms create new opportunities to interact with users. One of the available services now is personalized items recommendations. For providing these recommendations, various recommender systems algorithms were studied and developed in the last years. These algorithms also can be implemented in the game industry to recommend offers with game content inside the game. This opportunity is especially valuable for the games with the free-to-play business model, where players can play the game for free. However, they need to pay for additional content in the game. This paid content can create the most part of developers revenue and therefore offers should be shown to users reasonably. In our work, we analyse possible solutions for the creation of the recommender system for free-to-play games. We explore the specialities of the free-to-play business model, existing algorithms for recommender systems, common problems in this area and performance metrics of the algorithms. Then we test suitable recommender systems on the data from Facebook users of the social game “SuperCity” developed by Playkot company. For instance, we use several collaborative filtering algorithms such as variations of k-Nearest Neighbors, Singular Value Decomposition and clustering models. Finally, we discuss the empiric results and possibilities for future works. keywords: recommender systems, free-to-play, games, machine learning, collaborative filtering