Most of the existing data mining algorithms are processing in the centralized systems; however, at present large database is usually distributed. Compared with the frequent itemsets lost and high communication traffic in distributed database conventional and improved algorithm FDM, an improved distributed data mining algorithm LTDM based on association roles is proposed. LTDM algorithm introduces the mapping indicated array mechanism to keep the integrity of frequent itemsets and decrease the communication traffic. The experimental results prove the efficiency of the proposed algorithm.