A Method of Mining the Meta-association Rules for Dynamic Association Rule Based on the Model of AR-Markov

Based on the dynamic association rules, this paper puts forward the formal definition of meta-rules which makes use of the support vector and confidence vector as evaluation of rules, and introduces the usual mining process of the Meta-association Rules for dynamic association rule by the model of AR-Markov, the examples show that this method is effective in the analysing and predicting the change tendency of Meta-association Rules' support value and confidence value.

I. Introduction

Association rule is an important research in data mining and was first introduced by Agarwal and other researchers. This rule is used to find the relationships between items of the transaction data sets in order to provide reference for decision-making. The traditional algorithms of association rules consider that rules in the database is permanently effective. However, because the transaction data sets usually have the time characteristics, the rules change greatly with time. To set up a Meta-association rule for the rule is more intuitive to describe the tendency of rule changes and is better for decision-making. In order to describe the regularities of changes over time in association rules, Liu preliminarily introduced the new technology of dynamic association rules in reference [1]. This article, which is based on the fact that rules change with time, introduces the definition of Meta-rules of dynamic association rule, and uses integrated method such as AR-Markov models to mine the Meta-rules of dynamic association rule.