Research On The Algorithms Of Association Rule Mining

Data mining technique has attracted extensive attention in information technology industry. Association rule mining is one of major research directions in data mining and has tremendous applications in business, science and other domains.This thesis firstly reviews data mining technique, including the concepts of data mining, theoretical foundation of data mining and problems in data mining as well as systematic classification of data mining. Then, the algorithms of association rule mining is discussed, covering issues from Apriori algorithm to its improved algorithms. In order to make good use of the knowledge that has been mined to improve the mining efficiency when new data arrives, a new concept, sensitivity, is proposed. Then an improvement of the classical incremental updating algorithm (FUP) is presented according to the sensitivity. The two algorithms are compared and analyzed through an illustrating example. Experiments results show that the improved algorithm can find new interesting itemsets in the new dataset and has a better performance and sensitivity than FUP under different databases and different support thresholds.