Abstract: When the sequence database is updated，the existing incremental mining algorithms of sequential patterns only mention insert operation and append operation of sequences，and do not mention the delete operation.An incremental sequential patterns updating algorithm based on sequence tree is proposed，called ISPST.When the database is updated，ISPST only constructs the projected databases for the sequences associated with the deleted sequences，and completes the updated operations of the sequence tree，and finds all the sequential patterns through using depth-first search strategy to traverse the sequence tree.Experiments show that when the support is changed，ISPST outperforms PrefixSpan and IncSpan in time cost.

Key words: sequential patterns, incremental mining, projected database, sequence tree

摘要： 在序列数据库更新时，现有的增量式序列模式挖掘算法只提到序列的插入操作和序列的扩展操作两种情况，没有针对序列删除操作。提出了一种基于序列树的增量式序列模式更新算法（ISPST）。当数据库更新时，ISPST算法只需要对与删除序列有关的序列构造投影数据库，实现对序列树的更新操作，通过深度优先遍历序列树得到更新后数据库中的所有序列模式。实验结果表明，当支持度发生变化时，ISPST算法在时间性能上优于PrefixSpan算法和IncSpan算法。

关键词: 序列模式, 增量式挖掘, 投影数据库, 序列树