Mining Sequential Patterns with Pattern Constraint

Abstract

Mining sequential patterns is to find the sequential purchasing behaviors for most of the customers. There were many algorithms proposed for discovering all the sequential patterns. However, users may be only interested in certain items or behaviors. The items or patterns specified by users are called “pattern constraints.” If we first find all the sequential patterns and then filter out the patterns which the users are not interested in, then it will take much more time to find interesting sequential patterns. Therefore, the challenge for mining interesting sequential patterns is how to avoid searching for uninteresting sequential patterns, such that the mining time can be reduced. In this paper, we propose a query expression to represent the pattern constraints and an efficient mining algorithm to find sequential patterns which satisfy user specified pattern constraints. In our experiments, we compare our algorithm with well-known SPIRIT(R) algorithm on a real dataset. The experimental results show that our algorithm is more efficient than SPIRIT(R) algorithm.