Abstract—Book selection is an important operation in a library and should be conducted on the basis of reasonable data. This paper evaluates the process of book selection in a university library by analyzing the loan records. The authors classified the books in the library according to Nippon Decimal Classification (NDC), and investigated the turnover rate for each category. The results show that the book selection process in the library was appropriate in terms of the turnover rate although the process was not conducted with considering the turnover rate. Additionally, the authors considered the rate of the books that bought by researchers in the university and accessioned to the library. The results indicate some viewpoints for selecting books in university libraries. Index Terms—Data mining, library, book selection, loan record. I. INTRODUCTION Book selection is one of the important operations in a library. To make good use of public property, the way of book selection should be evaluated and improved according to the result of the evaluation. In some Japanese university libraries, however, a part of the process depends on the static guide independently set in each library or the individual discretions of a staff of the library. To meet the constantly changing demands of library patrons, book selection should be operated by a dynamic way according to objective statistics. There are two possible ways to evaluate the process of book selection, that is, evaluating the collection in themselves and analyzing use of the collection [1]. An analysis on use of books in a library had been conducted at latest 1961 [2]. Evans [3] investigated the numbers of loans for books classified by the person who had selected the book, academic staff, library staff, and book jobbers, in some academic libraries. Pritchard [4] also investigated the frequencies of loans for books selected on the basis of four criteria, new editions of already held books, requests on interlibrary loan, recommendations by library staff, and recommendations by academic staff. As for Japanese university libraries, Matsui and Isono [5] investigated the turnover rate of books in the library of Nara University, and found the typical change over the years (obsolescence) on the frequency of loans. The turnover rate is defined for a set of books to be the number of loans for a book. Koizumi [6] also Manuscript received May 5, 2014; revised July 25, 2014. Mitsuro Kitajima and Kensuke Baba are with the Library of Kyushu University, 10-1, Hakozaki 6, Higashi-ku, Fukuoka, 812-8581, Japan (e-mail: kitajima.mitsuro.023@m.kyushu-u.ac.jp, k.baba.060@m.kyushu-u.ac.jp). Toshiro Minami is with Kyushu Institute of Information Sciences, 3-1, Saifu 6, Dazaifu, Fukuoka, 818-0117, Japan (e-mail: minami@kiis.ac.jp). investigated the turnover rate of books in Keio University Library, and clarified the difference between book selections by librarians and faculties in the university. In our research, we considered the turnover rate for categorized books to find detailed tendencies in loan records. The aim of our study is to find useful information for book selection in a library from its loan records. An idea of a “good” book selection is that the selected book is well borrowed by patrons, that is, the turnover rate of the book is high. Additionally, from the viewpoint of collections for academic researches, the number of books and the turnover rate are better to be balanced for each research area. Computing the turnover rates for each book is technically easy, but it has the following practical difficulties:  Analyzing loan records in a library is often considered in connection with personal information or privacy matters,  The results of loan record analyses cannot be applied straightforwardly to selection of new books. The first difficulty seems to be remarkable especially in Japan because of historical reasons. The reason of the second difficulty is that book selection usually selects books that have not been already housed in the library. To avoid the difficulties, we classified books into some categories and identified the plural books in each category. In this paper, we evaluated the book selections in Kyushu University Library. In the library, the process of book selection is mainly conducted on the basis of  Requests from the faculties in the university (include designs in syllabuses),  Requests from patrons of the library, or  Selections by a staff of the library. Additionally, some books are bought by researchers from their funds, and such books are accessioned formally to the library, but physically to their laboratories. Currently, the loan records are not considered for the selections. We collected the book accession records from April 2000 to March 2013 and the book loan records from April 2012 to March 2013. We classified the books according to Nippon Decimal Classification (NDC), and investigated the relation between the rate of book accessions and the turnover rate for the books categorized into natural sciences. This investigation was also conducted for the book accession records from April 2012 to March 2013 to evaluate book selections in a short term. Additionally, we evaluated the book selections based on direct requests from researchers in the university in terms of the turnover rate. This evaluation was also conducted for books in social sciences, and the results showed a clear difference of the tendencies between the research areas. Our study suggests some viewpoints for improving book selection in university libraries rather than proposing the An Evaluation of Book Selection in a University Library by Loan Record Analysis Mitsuro Kitajima, Kensuke Baba, and Toshiro Minami International Journal of Information and Education Technology, Vol. 5, No. 10, October 2015 728 DOI: 10.7763/IJIET.2015.V5.601