The design of frequent sequence tree in incremental mining of sequential patterns

In the process of incremental mining, when the support is changed, the storage structure in existed incremental mining algorithms of sequential patterns determines that the algorithms need to mine the database once again. In this paper, we propose a novel data storage structure, called frequent sequence tree, and give the construction algorithm of frequent sequence tree, called Con_FST. The root node of the frequent sequence tree stores the frequent sequence tree support threshold and the path from the root node to any leaf node represents a sequential pattern in the database. Frequent sequence tree stores all the sequential patterns with its support that meet the frequent sequence tree support threshold, so when the support is changed, the algorithm which uses frequent sequence tree as the storage structure can find all the sequential patterns without mining the database once again. A pruning strategy is proposed to optimize the construction algorithm. Experiments show that the incremental mining algorithm of sequential patterns which uses the frequent sequence tree as the storage structure outperforms PrefixSpan in space cost.