Metadata Analysis for User and Video Game Purchases

Abstract

The video game industry has been continuously growing as technology integrated into our
everyday lives. Video game platforms such as Steam need to promote their games and be able to accurately
recommend relevant games to their users. This can be done through careful analysis on
User-Item data pairs along with various other fields on interest. We aim to explore both basic and complex
methods to determine if a user will purchase a game or not. Correctly predicting this could lead to various
benefits such as creating a recommender system that could predict the user rating or preference in relation to
a game.
By using naive approaches, we can achieve relatively simple baseline models that perform quite well.
However, we can see significant improvements with our complex models by extracting various features and
applying different approaches to the data