Author(s)

C Mooney & J F Roddick

Abstract

With the improvements in data warehousing and database technology, the amount of data being collected has grown at a remarkable rate. The field of data mining, which enables the extraction of interesting or relevant information from such data, has also grown at a similar rate. Methods now exist that enable the extraction of rules from varied data sources from which users are able to draw inferences about the underlying data. This paper surveys and extends a new area of data mining that has recently emerged – Rule Mining. Rule mining uses the results of previous mining sessions as input to a second mining process that produces rules with very different semantics which can be used to extend the inferences made from the underlying data. The paper first presents an overview of the work in this area. We then present an efficient method of longitudinal association rule mining that uses previously constructed frequent itemsets (rather than either the (larger) source datasets or the (larger) resultant rulesets) within a rule production engine. We also show how this method can be used for longitudinal association rule mining. 1 Introduction During the past decade the collection and storage of data has continued to increase at exponentially. The technology associated with warehousing and

Keywords