Optimize Recommendation System with Topic Modeling and Clustering

With the rapid development of e-commerce, recommender systems have been widely studied. Many recommendation algorithms utilize ratings and reviews information. However, as the number of users and items grows, these algorithms face the problems of sparsity and scalability. Those problems make it hard to extract useful information from a highly sparse rating matrix and to apply a trained model to larger datasets. In this paper, we aim at solving the sparsity and scalability problems using rating and review information. Three possible solutions for sparsity and scalability problems are concluded and a novel recommendation model called TCR which combines those three ideas are proposed. Experiments on real-world datasets show that our proposed method has better performance on top-N recommendation and has better scalability compared to the state-of-the-art models.