Mining frequent patterns in transactional databases is an important part of the association rule mining. Frequent pattern mining algorithms with single minsup leads to rare item problem. Instead of setting single minsup for all items, we have used multiple minimum supports to discover frequent patterns. In this research, we have used multiple item support tree (MIS-Tree for short) to mine frequent patterns and proposed algorithms that provide (1) a complete facility of multiple support tuning (MS Tuning), and (2) maintenance of MIS-tree with incremental update of database. In a recent study on the same problem, MIS-tree and CFPgrowth algorithm has been developed to find all frequent item sets as well as to maintain MS tuning with some restrictions. In this study, we have modified the maintenance method by adding the benefit of flexible MS tuning without any restriction. Again, since database is subject to practice, an incremental updating technique has been proposed for maintenance of the MIS-tree after the database is updated. This maintenance ensures that every time an incremental database is added to the original database, the tree can be kept in correct status without costly rescanning of the aggregated database. Experiments on both synthetic and real data sets demonstrate the effectiveness of our proposed approaches.
Download Full PDF Version (Non-Commercial Use)