The most important mission of libraries is to collect appropriate materials such as books and journals, and provide their patrons. In order to carry out this mission efficiently, libraries shelve them according to a classification system. The decimal classification (DC) system is the most popularly used one by libraries. For example in Japan, most libraries use Nippon Decimal Classification (NDC) system. However DC systems have a weakness in their flexibility for adjusting to such changes as the appearance of new concepts, rise and fall of popularity of existing subject fields, etc. For example, information science is relatively new for libraries, and it is hard to find an appropriate category code for it in NDC. Eventually, information science is assigned to the category code 007 in the third level of categorization of NDC. In this paper, we deal with an index for measuring similarity between NDC categories, which indicates how much amounts of the books of the comparing categories are commonly borrowed by library patrons. In this way, NDC categorization system can be tuned according to the current status of patrons' interest tendencies. Also, we demonstrate its usefulness by comparing the interest area profiles between the one using original NDC and the "virtual" NDC. By using the virtual NDC, we can recognize the practical interest area profiles of a patron or a group of patrons like faculties from loan record analysis. Studies in this direction will boost up advancing more useful library services toward the future.