The design of storage structure for sequence in incremental sequential patterns mining

The storage structure for sequence in existing incremental mining algorithms of sequential patterns is used to store the sequences that meet the support threshold. When the database is updated, the projected database needs to be constructed to find the sequential patterns in the updated database. In this paper, we propose the structure of sequence tree based on projected database, called sequence tree, and give the Stree_PS algorithm which is used to construct the sequence tree. Sequence tree is a novel data storage structure, it is similar in structure to the prefix tree. But the sequence tree stores all the sequences in the original database. The path from the root node to any leaf node represents a sequence in the database. The structural characteristic of sequence tree makes it suitable for incremental sequential pattern mining. Experiments show that the incremental mining algorithm of sequential patterns which uses the sequence tree as the storage structure for sequences outperforms PrefixSpan in space cost on condition that the support threshold is smaller.