Abstract: The proposed algorithm is based on the thinking of the frequent subtree mining repetitive DNA sequences in the body identified.The organization of DNA sequences in the new algorithm is different from with the others;organized a sequence as a tree, so we could avoid alignment as those traditional methods, then improved trees more simple that could be operating by frequent subtree mining, used a kind of algorithm for mining frequent subtree to learn these trees.This new algorithm could find out the repeated sequences which meet the threshold set directly, avoid the wasting of time result of splicing the short sequences.Designed the new technology "secondary identification", which could find out the fuzzy repetitive sequences, also improved integrity of identification.Experiment show that our mothod improved the time efficiency compared with mainstream algorithms, especially learning to find out some long sequences and highly comparable on the integrity of identification.