Research And Implementation Of Resource-adaptive News Recommender System

With the rapid development of Internet, the amount of the Web data is keeping increase. It is difficult for users to find the valuable information which attracts them in large-scale data. Additionally, there are 0.38 billion Web users in China, amounting for 29%, and more than 80% of Web users read Web news. Personalized recommendation systems are designed to provide personalized news to users. However, these systems are in primary stage in China.Firstly, available news recommendation systems are not good enough currently, since they can merely recommend news to users by asking users some questions. This paper proposes a recommendation method based on implicit feedback from user-generated profile. It gathers user behavior through client software and creates a multi-model user profile by analyzing these records. It considers both the short-term and long-term user interests and updates these interests automatically.Secondly, traditional researches of recommendation system mainly focus on recommending quality. However, the efficiency problem is very crucial in practical applications. This paper proposes a resource-adaptive algorithm to address this problem, which tries to balance the precision and the efficiency of the system. It monitors the free resource of the system and adjusts a sliding time window on news stream accordingly. This way, the dimensions of document vector and user profile vector change dynamically. When system load increases, the system can still reduce the computing time.Finally, this paper implements the algorithms and designs the news recommending system EagleNews. The system gathers news from famous news websites in China, and recommends news using the content-based method. Furthermore, it is a standard web service by using HTTP protocol and XML as its data exchange way. The results of experiments show that the performance of the systems is quite good.