Affiliations: [a] Department of Computer Science, New York Institute of Technology, New York, NY, USA | [b] Institute of Information Management, I-Shou University, Kaohsiung, Taiwan. E-mail: tcwang@isu.edu.tw | [c] Department of Electrical Engineering, National University of Kaohsiung, Kaohsiung, Taiwan. E-mail: tphong@nuk.edu.tw
Correspondence:
[*]
Corresponding author: Shyue-Liang Wang, Department of Computer Science, New York Institute of Technology, 1855 Broadway, New York, NY 10023, USA. Tel.: +1 212 261 1640; Fax: +1 212 261 1748; E-mail: slwang@nyit.edu.
Note: [1] Part of this work has been presented in the 2005 IEEE SMC conference.
Abstract: An Informative Rule Set (IRS) is the smallest subset of an association rule set such that it has the same prediction sequence by confidence priority [9]. The problem of maintenance of IRS is a process by which, given a transaction database and its IRS, when the database receives insertion, deletion, or modification, we wish to maintain the IRS as efficiently as possible. Based on the Fast UPdating technique (FUP) [5] for the updating of discovered association rules, we propose here two algorithms to update the discovered IRS when the database is updated by insertion and deletion respectively. Numerical comparisons with the non-incremental informative rule set approach show that our proposed techniques require less computation time, due to less database scanning and less number of candidate rules generated.
Keywords: Data mining, association rule, prediction, informative rule set, and incremental discovery
