Research On Quantitative Association Rules Model And Algorithm

Association rule mining being one of the key points in date mining field is to find such a rule as "80% customers who buy A and B also buy C and D" in a database with given minimum support and minimum confidence. In the early time, most of the studies base on transaction database consisting only of boolean attributes. When come to the databases in the science and business fields, attributes are not limited to being boolean but can be either quantitative or categorical. Mining association rules in such databases, called quantitative association rule mining, is a much more influential research problem when being compared to boolean association rules mining. The main content of this dissertation is as follow:(1) Model, algorithms and current researches on quantitative association rule mining are studied in this dissertation, and some known algorithms are analyzed and summarized, and the advantage and deficiency of these algorithms is represented, too.(2) Due to the combination of value intervals of quantitative attributes and computation of the support for those intevals costing lots of time and constituting a hamper toward the efficiency of the mining algorithm, algorithm BMIQAR mining quantitative association rules beased on attributes which have strong information relationship is proposed. The extended experiments show that the mining efficiency of algorithm BMIQAR has been improved greatly and most of the rules with high confidence are obtained.(3) As the existed studies just paid attention to either the weight or the quantitative value of the items only and rules mined using these methods didn't always accommodate to the requirements of the users when mining quantitative association rules in database which contain quantitative attributes, algorithm WQAR which conforms the weight and the quantitative value of the items and bases on the concept of K-profit support expectation is designed. Practical experiments show that algorithm WQAR can mine maximal profits itemsets and relationships between the itemsets.