Study Of Scientific Articles Recommendation Based On Hybrid Model

With the rapid development of Internet technology in recent years, the academic research filed has changed a lot, the number of academic papers on the network has increased explosively, and the academic resources information becomes complex and diverse. The researchers often need to spend a lot of time and efforts to find the valuable academic information on the Internet. Therefore, how to quickly and accurately find the academic information they are interested becomes an urgent problem.This paper mainly focuses on how to build the interesting model for the academic researchers and recommend scientific information to them accurately. Based on the study of topic model in content-based recommendation algorithm and model-based recommendation in collaborative filtering, we combine the two methods and propose a new hybrid recommendation method to improve data sparsity problem in collaborative filtering approach. In this paper, we implement the proposed method, determine the values of some parameters in the method, and compare the proposed method with other several approaches through the experiment. The experiment results verify the validity and superiority of the model.The method proposed in this paper includes a new topic model-ACTOT(Author Conference Topic Over Time) model which use the article content, the journal/conference information and published time of the scientific paper for accurately modeling the interest of researchers. The MFWT (Matrix Factorization With Topic)model merge content-based method and collaborative filtering together, and regularize the latent factor vector of user and paper in PMF (Probabilistic Matrix Factorization) model by the topic vector that computed by the ACTOT model and LDA model. The MFWT model improves the performance of PMF model, depresses the adverse effects caused by data sparsity effectively and also solves the cold start problem of collaborative filtering method.This article first analyzes the status and deficiency of academic research field, and then introduces the proposed MFWT model in detail, finally presents and analyzes the experiment results.