| With the popularity of mobile internet technology, the basic function of mobile client has been unable to meet more demand. More and more users have begun to become deep mobile internet user, which leads to a large number of mobile reading and news products. The online education news article is a kind of vertical reading field. The number of users in this field has exceeded 100 million. But for the providers of mobile news article service, it is a huge challenge to make the users to find what they are really interested in. This paper focuses on the personalized reading recommendation system in online education field.This paper first builds the data collection module of user reading behavior on the iOS platform through the seamless embedding technology. Then it exacts the article keyword to create VSM model through TF-IDF algorithm and reduces the dimension of the model through latent semantics analysis technology. Next it uses the user's reading behavior data to train the user's interest model through the Naive Bayes algorithm, which implement the content-based recommendation process. But the result of this recommendation technology is not ideal and leads to specialization problem. Therefore, this paper combines the collaborative filtering recommendation technology based on matrix decomposition with recommendation technology based on item popularity to build the whole recommendation engine of this paper to deal with various complex scenes. Compared with single-class recommendation algorithm, the recommendation engine is significantly improved.This paper also presents a design scheme of a complete personalized reading service system, and elaborates the design and implementation of each module, including the client business module, crawler and article entry module, backend server module and recommendation engine module. Finally, the online test of the system proves the effectiveness of the algorithm, and improves the user activity of the APP. |
PDF Full Text Request