In general, the recommendation system of the library is recommended based on the user’s preference, that is it is based on user behavior. It does not combine the user’s identity background and characteristics to recommend. Content-based recommendation main consider the internal information (That is MARC field of OPAC.) of the book as the assessment of whether a book should be recommended. However book’s content and introduction or other external information that we can get form MARC field has richer content features.Therefore, by the University Library and its users features, combined with the reader’s academic subject background. This thesis establishes user’s interest model based on user’s subject to recommend books which meet the information needs of users. And using text mining to establish book’s content features model which combine with MARC description information and other external information, such as:content, introduction and so on. From the user interest model and book feature extraction two aspects, trying to improve the efficiency of the book recommendation.This thesis uses the paired comparison (Pair Match) and hits (Hit Rate) as assessment indicator. The results show that recommendation model proposed in this thesis Enhance the effectiveness of book recommendation performance. The user interest model can enhance recommendation result’s hit rate, while binding books external information based on the recommended method of content can significantly enhance the recommendation result’s accuracy. That is significant improvement books recommend order. |