Abstract: To find all the frequent itemsets in a database rapidly, a new algorithm GMK is brought forward based on the database whose major key is goods. It records the ID that buys the itemset through an ID-array, then counts the number of elements that has the same ID in the itemset that is jointed to. Accordingly, the degree of support of the joint-item is obtained. GMK only scans the database once, thus it can reduce the quantity of calculation rapidly and can be realized easily. The validity of GMK is proved through tests.