User-association mining based on two-stage count

Both item-associations and user-associations mined from the rating table can be used to make personalized recommendation for the current user in rule-based recommend technique. Mining user-associations is the key for the recommendation based on user-associations. We find that the current user not only can be used to constrain the rule form in user-associations mining process, but also can be used to partition the rating table into two parts in order to accelerate user-associations mining. It is first proved that user-associations about the current user mined from the whole rating table are contained in those mined only from the data set that contain the current user´s rating. Then, a user-association mining frame based on two-stage count called TSCF is proposed. TSCF frame can be implemented by using existing algorithms for mining association rules. And an algorithm TSCF-CL for mining user-associations is implemented by using the concept lattice. Last the performance comparison with ASARM algorithm shows that TSCF-CL can reach better time capacity.