Mining interesting rules from a large set of association rules

Abstract

The goal of data mmmg IS to discover knowledge and reveal new, interesting and previously unknown information to the user. A central data-mining tool is association rules. This thesis work presents a constructive approach to mine interesting rules (CAMIR) from a large set of association rules. CAMIR first enumerates primitive rules with large interestingness and then generates rules with value grouping by merging interesting primitive rules. It assumes all attributes are nominal and there is no missing attribute value. Inputs of the algorithm are the number of attributes in bodies of rules, lower bound of support, accuracy and interestingness. The essence of CAMIR is that it requires minimum number of user specified parameters and users do not require any prior knowledge about the target problem. Here, the computation of interestingness of rules is simple and generalized. It has been applied on a number of benchmark data sets. They are mushroom, letter recognition, breast cancer, statlog (heart), hepatitis and liver disorders. The experimental result shows that CAMIR can return a large number of interesting rules within shorter interval of time in comparison with other existing algorithms.