Personalized Recommendation Model Based On Semantic Reasoning

Modern society has been brought many conveniences due to the rapid development of Internet technology. Along with the emergence of various types of Web sites, the huge amount of information brought by the Internet makes the research and application of data gradually become the trend of Internet development. The traditional "one-size-fits-all" model can not meet the needs of users of different backgrounds. It also have a big obstacle to the post operation of the website. The use of data, which is more popular in accordance with historical data for the user to personalize recommendation. There are some defects in the current main recommendation model:the importance of the relationship between ontology concepts is ignored, the user interest model updating is not complete, the user’s interest in the concept of interest can not be forgotten as time goes by. In order to solve these problems, this thesis studies the recommendation algorithm based on ontology, which is as follows:1. Based on ontology, a new user interest model is proposed, which is based on the theory of activation and the renewal of interest model and user’s interest. When the user’s interest is changed, the personalized recommendation system can update the entire user interest model based on the theory of activation. In the selection of the core algorithm, this thesis selects the mainstream K-means clustering algorithm and the hybrid collaborative filtering algorithm respectively to recommend, according to the actual situation to produce the recommended results.2. This thesis through the crawler to the two major authority website music and other related data capture, design and implementation of a personalized recommendation system based on user interest model. The system can analyze the data through the behavior data of historical users. Finally, it is recommended that users choose different kinds of music. Through a large number of experiments, the results show that the personalized recommendation system based on user interest model proposed in this thesis can be recommended for different users.