Majority Classification by Means of Association Rules

Associative classification is a well-known technique for structured data classification. Most previous work on associative classification based the assignment of the class label on a single classification rule. In this work we propose the assignment of the class label based on simple majority voting among a group of rules matching the test case. We propose a new algorithm, L3M, which is based on previously proposed algorithm L3. L3 performed a reduced amount of pruning, coupled with a two step classification process. L3M combines this approach with the use of multiple rules for data classification. The use of multiple rules, both during database coverage and classification, yields an improved accuracy.