Title: Partition-based sorted pre-fix tree construction using global list to mine maximal patterns with incremental and interactive mining

Authors: R. Vishnu Priya; A. Vadivel

Addresses: Department of Computer Applications, National Institute of Technology, Tamilnadu, India ' Department of Computer Applications, National Institute of Technology, Tamilnadu, India

Abstract: Maximal pattern mining in highly dynamic transactional database is difficult task since both discarded and updated contents are used together by changing the threshold. This essence has spurred the researchers to develop algorithm to support both incremental and interactive mining, which do not identify the pattern again and again for a corresponding update in the database. In this paper, a partition-based sorted prefix-tree (PSP-tree) is proposed and is constructed in two phases with single scan. Initially, the sub trees are built along with its item lists and are merged to obtain the global sorted item list. Later, each node of the sub trees is dynamically rearranged based on the sorted list and merged to form the proposed tree to support incremental and interactive mining. Performance of PSP-tree is encouraging especially for sparse database and the patterns are mined very fast compared to recent similar approach.

Keywords: maximal pattern mining; association rule mining; partition strategy; partition-based sorted prefix tree; PSP tree; knowledge discovery; incremental mining; interactive mining; sparse database.

DOI: 10.1504/IJKEDM.2012.051238

International Journal of Knowledge Engineering and Data Mining, 2012 Vol.2 No.2/3, pp.137 - 159

Published online: 13 Sep 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article