Research On The Application Of University Library Personalized Recommendation Service Based On Data Mining

As the categories and quantity of the goods are increasing quickly, the online shopping platforms applied the recommendation systems in succession, which implemented recommending goods that the users may interested in according the users’ historical purchase records and browsing histories on the websites. And the university library books are increasing fast too, readers will spend more and more time to borrow books from library if they don’t have specific lending targets. As for this situation, this paper raised the research of the application of content-based recommendation algorithm in the university library personalized recommendation service. By learning the experience of the online shopping platforms, this paper aimed to establish personalized book recommend systems for university libraries, and to recommend books to readers that they may interested in initiatively, by analyzing the readers’ historical borrowed records and browsing histories.Taking Inner Mongolia University of Technology library as an example, and using My Eclipse as development tool, this paper researched the application of content-based recommendation algorithm in the university library personalized recommendation service in detail. Firstly, it used data cleaning, data conversion and data normalization, three data preprocessing methods to preprocess the original data, which made the data more standardized and easier to find the rules. After preprocessing the data, it used the tokenizer ICTCLAS of Chinese Academy of Sciences to do the segmentation of books. Then it used method TF-IDF to calculate the weight of each word, and the words that with higher weights can be the key words to denote the main feature of the books. Then it constructed the vector space model of all the books and their key words. At last, it computed the similarity between every two books by using the VSM, and then sorted the books that have higher similarities with the books that the readers once borrowed from the library and recommending them to readers. In order to make the recommendation results diverse, it also recommended books by combining with book publishers, the authors and types information.In the process of research, the solutions to the main problems are: The Chinese word segmentation procedure was adjusted by improving the user dictionary and increasing the stop words, which made the accuracy of the segmentation result higher, and made it more suitable for the segmentation of books’ names. A solution for the application of short texts of words weighting algorithm TF-IDF was given which made it more suitable, and more fair and reasonable for book names. And it raised a solution with triple to the problem of sparse matrix that caused by vector space model constructing.At last, this paper evaluated the recommendation results by using the mature evaluation methods. In order to make the recommendation results more visualized, and make it easier for readers to use, the system interface was designed, and the recommendation results was displayed on the interface in the form of a list. After each recommendation result, there was its recommendation reason, which was clear and could easily accepted by readers.