Recommender System: Rating predictions of Steam Games Based on Genre and Topic Modelling

摘要

In this modern society, the majority of e-commerce platform have a recommender system. Recommender system is a popular and powerful way to introduce users with suggestions that they are most probably going to buy or use. The research conducted mainly focuses on implementation of genre-based and topic modeling model in a recommender system to predict rating of games for a user using a public Steam dataset. Both models will also be combined to implement a hybrid recommender system. Our models use KNN algorithm to predict rating of a targeted user. The system is fully implemented in Python programming language. Multiple Python libraries were utilized for data cleaning process. All predicted ratings generated were evaluated and compared to each other. Based on results evaluated, genre-based model outperforms both topic modeling and hybrid models. However, the performance of genre-based model doesn’t outperform the model performance from previous research. Therefore, it can be concluded that genre isn’t a suitable parameter for recommending games.