Sequential Association Rule Mining with Time Lags

Sherri K. Harms, Jitender S. Deogun

Research output: Contribution to journalArticle

87 Scopus citations

Abstract

This paper presents MOWCATL, an efficient method for mining frequent association rules from multiple sequential data sets. Our goal is to find patterns in one or more sequences that precede the occurrence of patterns in other sequences. Recent work has highlighted the importance of using constraints to focus the mining process on the association rules relevant to the user. To refine the data mining process, this approach introduces the use of separate antecedent and consequent inclusion constraints, in addition to the traditional frequency and support constraints in sequential data mining. Moreover, separate antecedent and consequent maximum window widths are used to specify the antecedent and consequent patterns that are separated by either a maximal width time lag or a fixed width time lag. Multiple time series drought risk management data are used to show that our approach can be effectively employed in real-life problems. This approach is compared to existing methods to show how they complement each other to discover associations in the drought risk management domain. The experimental results validate the superior performance of our method for efficiently finding relationships between global climatic episodes and local drought conditions. Both the maximal and fixed width time lags are shown to be useful when finding interesting associations.

Original languageEnglish (US)
Pages (from-to)7-22
Number of pages16
JournalJournal of Intelligent Information Systems
Volume22
Issue number1
DOIs
StatePublished - Jan 2004

Keywords

  • Drought risk management
  • Knowledge discovery
  • Sequential rule discovery
  • Time lag

ASJC Scopus subject areas

  • Software
  • Information Systems
  • Hardware and Architecture
  • Computer Networks and Communications
  • Artificial Intelligence

Fingerprint Dive into the research topics of 'Sequential Association Rule Mining with Time Lags'. Together they form a unique fingerprint.

  • Cite this