ABSTRACT
The model-based collaborative filtering recommendation took advantages of the training dataset to build a model of user behaviour, then predict the ratings. Such methods are mostly related to machine learning, which mainly includes the method based on linear algebra, such as the SVD (Singular Value Decomposition) (Peterek A., et al., 2007), PCA (Principal Component Analysis) (Goldberg K., et al., 2001), Bayesian networks (Sum X. Y., et al., 2006, etc. Model-based collaborative filtering algorithms can improve the sparsity of memory-based collaborative filtering algorithms to some extent, but they face the problem of serious complexity. They need a relatively complex process to solve the parameters of the training model and predicting according to the model. Since they need more resources, this results in the problem of algorithm expansibility.
