Binary Information Press
It is very time-consuming to mine sequential patterns from scratch each time when database change. Many sequential pattern mining data structures follow a complex hash-tree structure which has poor locality. In this paper, an efficient incremental algorithm called ISC (Incremental Sequential pattern mining based on Constraints) is proposed. This algorithm uses lattice structure to store the previous mining results and transforms dataset into a Bit-string map with scanning database only once. Compared with SPAM algorithm, the required memory size for storing temporal data during mining process of the authors' method is less than the one needed by SPAM algorithm when handling the original database.