Mining association rules in non-transactional databases
- Authors
- Lee, Ho-Jong; Lim, Seung-Hwan; Oh, Hyun-Kyo; Cho, Jinsoo; Kim, Sang-Wook; Cha, Jaehyuk; Lee, Junghoon; Kim, Hanil
- Issue Date
- Nov-2012
- Publisher
- INT INFORMATION INST
- Keywords
- Data mining; association rule mining; non-transactional databases; clustering; characterization
- Citation
- INFORMATION-AN INTERNATIONAL INTERDISCIPLINARY JOURNAL, v.15, no.11B, pp.5055 - 5069
- Indexed
- SCIE
SCOPUS
- Journal Title
- INFORMATION-AN INTERNATIONAL INTERDISCIPLINARY JOURNAL
- Volume
- 15
- Number
- 11B
- Start Page
- 5055
- End Page
- 5069
- URI
- https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/164328
- ISSN
- 1343-4500
- Abstract
- Association rule mining typically targets transactional data. In order to process non-transaction data in association rule mining, interval identification for each attribute is required. Previous methods perform the interval identification and rule mining steps independently and therefore cannot reflect the change of confidence according to the interval change in discovering association rules. This leads to improper identification of the intervals of attributes, thereby making association rules of high levels of confidence missed in a mining result. In this paper, we propose a novel method to identify good intervals of attributes by performing interval identification and rule mining steps simultaneously. The proposed method adjusts the intervals of attributes during performing the two steps. This makes it possible to find good intervals, and thereby discovering more association rules of high levels of confidence. The proposed method employs hierarchical clustering to group the attributes on the right hand side (RHS) of a rule, and also performs characterization analysis to assess the goodness of each cluster. According to our experimental results with real-world data, the proposed method results in finding useful association rules more than previous methods. Also, the level of confidence of the rules found by our method is greater than those by previous methods.
- Files in This Item
- There are no files associated with this item.
- Appears in
Collections - 서울 공과대학 > 서울 컴퓨터소프트웨어학부 > 1. Journal Articles
![qrcode](https://api.qrserver.com/v1/create-qr-code/?size=55x55&data=https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/164328)
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.