Application Of Generalized Association Rule Mining In The Library New Book Recommendation

Recommending new books plays an important role in providing personalization information service for readers which changes the passive service means to an active one by recommending new books according to reader's profile.After investigating conventional Generalized Association Rule mining algorithms, it is found that these algorithms are unable to satisfy the demand of association rule mining for the library new book recommendation. Therefore, a BASIC algorithm based on the MMS_Cumulate and GP-Apriori algorithm according to the characteristics of library new book recommendation is proposed first which is not very fast. Then based on the BASIC algorithm, a novel algorithm is presented, namely, MAR_LCR which is capable of finding generalized association rules in the form of "patron-book" and allowing users to specify multiple minimum supports to different items. Synthetic dataset is used to evaluate the performance of MAR_LCR algorithm. Experimental result shows that the algorithm is very effective. A discussion is made on how to identify relevant reader attributes in regard of the books they checked out and MAR_LCR algorithm is applied to the circulation records to mine the association rules. Finally, interestingness of rules is defined and an algorithm for pruning uninteresting rules is proposed.