Combined approach for mining high quality interesting association rules

Abstract Association rule mining is an essential data mining job used for mining relationships present within the data. Support - confidence framework is traditionally used for mining Association Rules. But the limitation is huge number of Association Rules are generated, which may be uninterested to the user. So, measures used for filtering Association Rules have been proposed in literature. They are classified into Subjective & objective. Subjective measures takes into account the user’s knowledge and his/her interest in pattern analysis, where as objective measures depends on structure of the pattern and the input data. Despite of advantages of subjective and objective measures, both the measures has limitations. In case of subjective measures, user’s can feel difficulty in expressing interestingness during pattern evaluation, where as in some cases objective measures may not generate really interesting rules. Hence an approach is proposed which generates interesting high quality association rules, which takes into account the advantages of existing measures.