Abstract: The efficiency of mining association rules is an important field of Knowledge Discovery in Databases. In this paper we have proposed an improved AprioriTid algorithm with transactions reduction, candidate itemsets reduction and support matrix to solve the bottleneck of itemsets generation. The highly efficient method described in this paper minimizes the database by deleting many transactions which need not be scanned. We also show a method to reduce the number of candidate itemsets by optimizing the join procedure of frequent itemsets and a support matrix method to accelerate the verification speed of candidate itemsets is put forward. To this end, the IAT algorithm for mining frequent itemsets, which is the improvement algorithm of AprioriTid, is designed in this article. The experiment results of the algorithm show that the improved algorithm can decrease related computation quantity in large scale and improve the efficiency of the algorithm. The simulation results of knowledge acquisition for fault diagnosis also show the validity of IAT algorithm.