Abstract: Aiming at how to mine frequent itemsets from affair database after specifying customer characters, the conception of mining with multidimensional constrained is brought forward.The FP_Growth algorithm is employed, but the way for constructing FP_Tree is different in every item-node, it's not the count of the item occurring in the database, but using a vector that makes of counts of the item under every constructing, the node vector and level structure are saved in HASH table, it can support mining the frequent itemsets in facultative constructing without scanning the database again.The result of experiment proves this algorithm is self-contained and has higher efficiency by comparing to what mine after selecting.