Artykuł w czasopiśmie
Brak miniatury
Licencja
ClosedAccessDostęp zamknięty
 

Efficient method for updating class association rules in dynamic datasets with record deletion

Uproszczony widok
dc.abstract.enAssociation rule mining is an important topic in data mining. The problem is to discover all (or almost all) associations among items in the transaction database that satisfy some user-specified constraints. Usually, the constraints are related to minimal support and minimal confidence. Class association rules (CARs) are a special type of association rules that can be applied for classification problem. Previous research showed that classification based on association rules has higher accuracy than can be achieved with an inductive learning algorithm or C4.5. As such, many methods have been proposed for mining CARs, although these use batch processing. However, datasets are often changed, with records added or/and deleted, and consequently updating CARs is a challenging problem. This paper proposes an efficient method for updating CARs when records are deleted. First, we use an MECR-tree to store nodes for the original dataset. The information in the nodes of this tree are updated based on the deleted records. Second, the concept of pre-large itemsets is used to avoid rescanning the original dataset. Finally, we propose an algorithm to efficiently update and generate CARs. We also analyze the time complexity to show the efficiency of our proposed algorithm. The experimental results show that the proposed method outperforms mining CARs from the dataset after record deletion.
dc.affiliationUniwersytet Warszawski
dc.contributor.authorNGUYEN, NGOC THANH
dc.contributor.authorNguyen, Hung Son
dc.contributor.authorNguyen, Loan T.
dc.contributor.authorVo, Bay
dc.date.accessioned2024-01-24T22:31:40Z
dc.date.available2024-01-24T22:31:40Z
dc.date.issued2018
dc.description.financeNie dotyczy
dc.description.number6
dc.description.volume48
dc.identifier.doi10.1007/S10489-017-1023-Z
dc.identifier.issn0924-669X
dc.identifier.urihttps://repozytorium.uw.edu.pl//handle/item/105872
dc.identifier.weblinkhttp://dx.doi.org/10.1007/s10489-017-1023-z
dc.languageeng
dc.pbn.affiliationcomputer and information sciences
dc.relation.ispartofApplied Intelligence
dc.relation.pages1491–1505
dc.rightsClosedAccess
dc.sciencecloudnosend
dc.subject.enData mining
dc.subject.enClass association rules
dc.subject.enRecord deletion
dc.subject.enUpdating rules
dc.subject.pleksploracja danych
dc.titleEfficient method for updating class association rules in dynamic datasets with record deletion
dc.typeJournalArticle
dspace.entity.typePublication