Incremental update strategy for indexed item set mining

Most of the research activities in association rule mining focuses on defining efficient algorithms for item set extraction. To reduce the computational complexity of item set extraction, support constraint is enforced on the extracted item sets. Recent existing work, IMine index (Item set-Mine index), a data structure, provides a compact representation of transactional data supporting efficient item set extraction from a relational DBMS. However, when the transactional database is updated, IMine index needs to be rematerialized. The proposed work presents an incremental update strategy to work on the dynamic transaction of DMBS for efficient item set extraction. Since no support threshold is enforced during the index creation phase, the incremental update is feasible without accessing the original transactional database. The index performance in terms of incremental updates is experimentally evaluated with data sets characterized by different size and data distribution. The execution time of frequent item set extraction based on incremental update strategy of IMine is better than the state-of-the-art algorithm i.e., existing IMine algorithm without update strategy. The experimental result shows the scalability of incremental update strategy for more frequent database updates characterized by a large number of transactions and with different pattern lengths.