Mining Positive and Negative Association Rules Based on Closed Itemsets: An approach for generalized rules

摘要

Association rules are a way of finding interesting patterns in business databases. We have considered the problem of finding positive and negative association rules based on the closed itemsets. Negative rules are also very useful in association analysis. These rules are useful in market-basket analysis to find the objects that are in clash with each other. There are two types of negative association being classified as general rules and confined rules. Confined negative association rules are of the type where all the items of antecedent(X) or items of consequent(Y) are either negated or non-negated e.g. ～(ab) => c or ～a and ～b => c which means negation of both items a and b implies c. In these types of rules, all items of antecedent or consequent are in conjunction of either negated or non-negated elements. Generalized association rules are of the type where an item of X or Y can be negated (i.e. a rule for which its antecedent or its consequent can be formed by a conjunction of presence or absence of items) e.g. ～a and b => c; a and ～b => c; ～a and ～b =>c, these can be three cases of ～X => Y. In the most of the research papers, we have hardly found any work done on the generalized association rules. We are dealing with generalized association rule. Usually mining for such rules requires the examination of an exponentially large search space. Despite of their significance, we have found very few algorithms to mine negative association rules. In this paper, the focus is to mine positive and negative association rules based on closed itemsets, which do not need the examination of exponentially large search space.