Abstract: To solve the problemthatthe existing algorithms ofmining association rules result in a numberof rules, upper closed set of an item set and generalized association rule base were defined. And some important propositions were obtained,e.g.the upper closed set of an item set exists uniquely and is the item set. From these propositions, the generalized association rule base of a transaction database was proposed. It contains the information of all the association rules in the database and the rules have a uniform structure. As a result, the number of rules can be reduced as many as possible. In addition, a mining algorithm for generalized association rule bases was proposed and its validation was proven using an example.