An Improved MFP-Miner Mining Algorithm and Its Application in Book Mining

Abstract

With purchase of new books every year, collection of books in a library will continuously increase. As circulation data is updated every day, database of a library will also increase while minimal support degree will change due to different required correlation degree. To satisfy requirements of library readers as soon as possible with book allocation, mining result needs to be updated continually. Therefore, this thesis proposes an IM-Miner algorithm to realize synthesized update and mining of maximal frequent item sets with database and minimal support degree varying at the same time. IM-Miner algorithm makes the most of FP-Tree features and doesn’t need to generate maximal frequent candidate item sets during mining. As generation of maximal frequent item sets only occurs in FP-Tree, no scanning of transactional database is required. Experimental results showed that IM-Miner algorithm is more efficient than other algorithms.