Abstract: To further reduce both CPU and I/O overhead in the process of mining the association rules on the large transaction database by the traditional algorithm, an improved algorithm of association rule mining based on vertical data layout named VARMLDb(Vertical Association Rule Mining for Large Databases) was suggested. In the proposed algorithm,after dividing the database into several partitions each of that is suitable for the current memory, the algorithm combines directed acyclic graphs and diffset(difference of tidlist sets) which belongs vertical data layout structure for storing and computing frequent item sets, which not only greatly cuts down the required memory size used to save intermediate results but also solves the low efficiency problem during the mining dense database by traditional vertical data mining algorithm, so that the algorithm is more effective for large dense databases. As a result of drawing the advantages of CARMA(continuous association rule mining) algorithm, the algorithm needs to scan the database for only twice.Experimental results show that the algorithm is correct, and in the large dense transaction databases, VARMI_Db algorithm has higher implementation efficiency. Continuous association rule mining algorithm, Directed acyclic graphs, Diffset plumb, Vertically distributed data, Dense database

Key words: Continuous association rule mining algorithm, Directed acyclic graphs, Diffset plumb, Vertically distributed data, Dense databases