Mining of association rules is an important problem in data mining, given a large set of data, extracting frequent item sets in this set is a challenging job in data mining. Item sets matching is the chief problem in extracting frequent item sets. And item set matching is the bottleneck of the mining process. It also has been proved that extracting frequent free item sets is a useful method. Many efficient algorithms have been proposed in the literature. The idea presented in this paper is to divide the database into multiple partitions and then find frequent free item sets in each partition, then merge the several partitions to generate other frequent free item sets and count the support. The algorithm costs little memory to save additional support numbers of item sets in each partition but greatly reduces the time of item set matching which is the bottleneck of the mining process. The experiments on real datasets have showed its good performance.