A novel algorithm for extracting knowledge based on mining multi-level sequential patterns

In this paper, we consider the problem of finding sequential patterns in multi-levels, with aid of candidate generate and test approach. Based on this technique, many sequential pattern algorithms have been developed, but few of them have paid attention to multi-level concept. We thus purpose a new efficient algorithm, called dynamic vertices levelwise (DVlw) for mining multi-level sequence patterns. It uses the same principals as other candidate generation and test algorithms but handles multi-levelled property for sequences prior to and separately from the testing and counting steps of candidate sequences. Empirical evaluation using synthetic data indicates that the proposed algorithm performs significantly faster than a state-of-the-art algorithm with this approach.