At present, the existed incremental mining algorithms can not make full use of the results of the previous mining. When the support is changed, the algorithms need to mine the database once again. In this paper, we propose an incremental mining algorithm of sequential patterns based on frequent sequence tree, called IMFST. When the database is updated and the support is changed, IMFST is divided into four kinds of situations to update the frequent sequence tree, and finally gets all sequential patterns. IMFST uses two kinds of pruning strategies to reduece the size of the projected databases. When the support is changed and the support is no less than the frequent sequence tree support threshold, IMFST can find all the sequential patterns without mining the database once again. Experiments show that IMFST outperforms PrefixSpan and IncSpan in time cost.