ARTMAP-Based Data Mining Approach and Its Application to Library Book Recommendation

Abstract

To overcome some disadvantages of the conventional data mining methods, a model based-approach to data mining by using supervised ARTMAP neural network is proposed and applied to a library book recommendation system. The proposed algorithm is based on formation of reference vectors that make a data mining system able to classify user profile patterns into classes of similar profiles, which forms the basis of a library book recommendation system. A correspondent computer program is developed by using C++ language. To evaluate the performance of the presented approach, the book circulation data of a university library is collected and used for the developed program. Simulation experiment results show that the ARTMAP network provides better performance than both ART2 network and the popular memory-based neighborhood algorithm.