Based on access sequence for deep learning model of news recommendation

• 1、Beijing Key Laboratory of Intelligent Telecommunications Software and Multimedia (Beijing University of Posts and Telecommunications), Beijing 100876

Abstract：In recent years, personalized news recommendation is an extremely challenging problem because of the rapid changes in user preferences and news characteristics. In order to generate recommendation better, the news recommendation system based on the deep reinforcement learning model has been rapidly developed. But deep reinforcement learning models still have some problems. First, it has inefficient sampling efficiency and is very hard to design a good reward function. Second, few models consider the user access sequence in news recommendation systems based on deep reinforcement learning model. These problems make the recommendation efficiency poor. In this paper,deep learning and user access sequence factor are combined to redesign the scoring standard of the items to be recommended. Moreover, this paper incorporate the public\'s preference for users with sparse access sequences. Finally, this paper has conducted experiments on some datasets provided by the NewsREEL platform and extracted from NetEase News. The results demonstrate that our proposed model can effectively improve the quality of the recommendation.

Keywords： news recommendation deep learning access sequence