Research On Relevant Book Recommendation Technology Based On Content

Collaborative filtering is one of the most popular technology used in e-commerce recommending system currently. But the fatal shortcoming of collaborative filtering technology is that it must have to know the information of user interests to make personal recommendation. And these information are difficult to obtain in the website with small users. So, in this paper we try to convert the research perspective, place extra emphasis on mining the relationship between items, and then recommend correlative items based on these relationships without considering the actual user. In this paper, we take books as our study object, take into account the interconnected characteristic of books by content similarity, and then make related recommendation of books. When the user searches or browses the book on the web sites, the correlative books will also be provided. The main research contents of this paper include following 4 parts:Firstly, considering the e-commerce recommending system cannot provide the whole content of books, we present a new recommending method which is based on the similarity of attributes among books. Based on this idea, we give our first book recommending strategy. In this way, the system is inclined to recommend books which have similar content with original book, so users can choose the perfect one suited to their own tastes among these books.Secondly, focused on the issue of some users are prefer to books which have correlative but not very similar content with original book, we adopt a measurement method of document correlation based on term correlation, and make recommendation using this way. This recommending method tends to recommend books which have relativity with the original book, so the users can get more relative information when they are searching or browsing target books.Thirdly, in order to improve the speed of recommending system when facing massive data, we propose to divide the whole book collection into clusters using document clustering technology, and then make the recommendation just in its own cluster. This method reduced the complexity of algorithm.Fourthly, we implement an on-line book recommendation system using the methods proposed in this paper. And this system will be attached to the knowledge service platform of haitianyuan,When the user search a book in the system, related books will be recommended to each book in the search results.