Abstract:Data mining means extracting or "mining" knowledge from large amounts of data to achieve the effective utilization of data resources.The association analysis is one of the widely used functions of data mining.It is used to locate association rules which describe the interesting connections among the items of a given data set. The thesis puts forth a genetic-based mining algorithm of association rules.In terms of coding,a method is presented to encode the individual with variant length so as to reduce the amount of the invalid rules to some extent.In terms of the construction of adaptive functions,some improvements are made.And in terms of the overlap of the rule library,owing to the inadequate data available for mining,flag bits which mark the mined rules are employed to avoid the repetition of rule mining.