Complementary Occurrence and Disjunctive Rules for Market Basket Analysis in Data Mining

H.D. Kim (USA)


Data Mining, Market Basket Analysis,Association Rules, Complementary Occurrence.


We present a scheme by which we can build disjunctive association rules for representing complementary occurrence of product items, as opposed to co-occurrence of items targeted under the current Market Basket Analysis. A disjunctive rule comprises a set of items that occurs in a mutually exclusive way. For example, "if coffee then regular or decaf" is a disjunctive rule, which says a person drinks or buys either regular or decaf coffee but not both. Although this type of association rule is not recognized in the current Market Basket Analysis, we believe it can be of great value for planning and analysis, and provide interesting observations on customer behavior. For one thing, we can get the list of competing items at a glance along with confidence rate for each item, based on transactions not on total sales figures. Our method for building a disjunctive rule consists of collecting and verifying complementary items from major items to minor items, and then successively adding them to the rule until the desired level of confidence is reached. One major advantage of this method is it can be easily added to existing Market Basket Analysis techniques without causing much changes or extra overhead, for example, as part of post-processing analysis.

Important Links:

Go Back