In this thesis, we propose a data mining approach to recommending library new books that have never been rated or borrowed by users. In our problem context, users are characterized by their demographic attributes, and concept hierarchies can be defined for some of these demographic attributes. Books are assigned to the base categories of taxonomy. The proposed approach starts with the identification of the type of users who are interested in some specific type of books. We call such knowledge generalized profile association rules. Less interesting or redundant generalized profile association rules are then pruned to form a concise rule set. The resultant rule set is then used for promotion of new books. We develop a new definition of rule interestingness with respect to book recommendation, propose an approximation scheme for estimating the interestingness of a rule, and construct a scheme to effectively conduct new book recommendation by using the interesting rules. We finally apply the book circulation data of a university library to the proposed approach for performance evaluation.