Audit Oriented Fast Algorithm for Sequential Mining with Constraints

Sequence pattern mining is one of important problems in data mining. Many algorithms have been proposed to solve this problem. But most of them consider little about time that is the main feature of sequential data. In this paper we present a fast algorithm SPIC (sequential pattern mining with constraints) based on SPADE. It utilizes time and attribute-relative features to lead the sequential mining process. And it uses key attribute to prune useless or misleading sequential rules. Finally, experiments on a real-world audit dataset show SPIC outperforms SPADE, especially when the number of attributes is far less than the average number of attribute values