Improvements of IncSpan: Incremental Mining of Sequential Patterns in Large Database

In reality, sequence databases are updated incrementally. The changes on the database may invalidate some existing sequential patterns and introduce new ones. Instead of recomputing the database each time, the incremental mining algorithms target efficiently maintaining the sequential patterns in the dynamically changing database. Recently, a new incremental mining algorithm, called IncSpan was proposed at the International Conference on Knowledge Discovery and Data Mining (KDD’04). However, we find that in general, IncSpan fails to mine the complete set of sequential patterns from an updated database. In this paper, we clarify this weakness by proving the incorrectness of the basic properties in the IncSpan algorithm. Also, we rectify the observed shortcomings by giving our solution.