To discover frequent itemsets beginning with each data item in CanTree, we propose GTree (Group Tree) as a projection- tree that is constructed from the base-tree. A node of GTree represents a set of nodes that have the same data item in
the base-tree (CanTree or parent GTree). Our algorithm, called CanTree-GTree, is very simple and efficient. This is because firstly it sorts the data items of a transaction alphabetically, then just adds them to CanTree. Secondly, it uses GTree to discover frequent itemsets by an efficient algorithm, the construction cost of which is lower than other data structures, such as FPTree and CanTree