الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 86 |  |  |
| | | from | 86 | | |
Abstract
The problem of mining association rules has attracted a lot of attention in the research •Community. Several techniques for efficient discovery of association rules have appeared. However, it is nontrivial to perform incremental mining or efficient mining of constrained association rules, in spite of their practical benefits. The research community has recently focused on providing separate solutions for these two problems.
Since many believe that constrained mining will be the standard, incremental mining of constrained rules will be necessary. In this thesis, a new algorithm, ICAP, for incremental mining of constrained association rules is introduced. The concept of constrained negative border is also introduced and its usage for the maintenance of constrained association rules when new transaction data is added to a transactional database is proposed. This fast incremental mining technique is applied to the constrained associar6n mining algorithm CAP. A key feature of the proposed algorithm is that it requires ~ most one scan of the original database only if the database insertions cause the constrain2d negative border to expand. Thus the speed-up of incremental mining is combined with the flexibility of the framework of constrained frequent-set queries, CFQ<~, used for specifying user constraints on the produced association rules. The correctness ofthe proposed algorithm is studied and a proof of it is given.
Several experiments were conducted to measure the relative performance of the new algorithm compared to the single alternative approach available so far, which is rerunning the CAP algorithm on the whole updated database. The results of the experiments show a significant improvement over rerunning CAP in almost all ofthe cases.