Full Citation

Title: A Temporal Extension of Closed Item Sets for Change Mining

Citation Type: Miscellaneous

Publication Year: 2008

ISBN:

ISSN:

DOI:

NSFID:

PMCID:

PMID:

Abstract: Frequent pattern mining often produces a vast set of results. To overcome this problem, two fundamental approaches are commonly employed: condensed representations, such as closed itemsets, and relevance assessment. In recent years, the change of itemsets over time is gaining increasing attention as a promising basis for developing novel, more comprehensible relevance assessment methods. One of the unsolved problems is that typically many of the observed changes are the side-effect of other changes. Existing condensed representation approaches fail in removing such redundancies because they have not been developed with the temporal dimension in mind. This paper proposes a novel approach for a condensed representation of itemsets which is based on utilizing temporal redundancies. In particular we prove that our approach yields a temporally non-redundant subset of closed itemsets which we therefore call temporally closed itemsets. Our experiments with real-life data sets show that the set of temporally closed itemsets is significantly smaller than the set of closed itemsets.

User Submitted?: No

Authors: Spott, Martin; Kruse, Rudolf; Bttcher, Mirko

Publisher: Otto-von-Guericke-Universitt Magdeburg

Data Collections: IPUMS USA

Topics: Methodology and Data Collection, Other

Countries: