Laws on Support Counts of Apriori Algorithm Candidates

This paper proposes three novel laws for finding useful candidates in database and preventing useless candidates by researching frequent itemset support. Some new concepts are introduced so as to explain these three laws. For example, independent itemset and support count, which can effectively avoid losing any interest associational rules during pruning, are introduced. In this paper, firstly, some key definitions on these approaches are presented such as support count, constant support count on fixed item sets then three novel approaches are detailed. Secondly, demonstrations and applications on these approaches are presented. The application of these approaches is used to test significances of these laws by an example which proves that these novel approaches can more efficiently reduce redundant candidates than A-priori algorithm.