Feature extraction of gameplays for similarity calculation in gameplay recommendation

This paper proposes a method for extraction of relevant features that represent a gameplay and are needed in gameplay recommendation. In our work, content based filtering (CBF) is adopted as the recommender algorithm. CBF exploits a heuristic that the user's previous ratings of items, gameplay clips in our case, can be used to derive the rating of an unrated similar item. In this work, in order to calculate the similarity between a pair of gameplays, a kind of autoencoder called Denoising Autoencoder is employed. Our experimental results confirm that the method can successfully extract features, based on which the resulting similarity between a pair of gameplays matches with their content and human perception.