TKS: Efficient Mining of Top-K Sequential Patterns

Sequential pattern mining is a well-studied data mining task with wide applications. However, fine-tuning the minsup parameter of sequential pattern mining algorithms to generate enough patterns is difficult and timeconsuming. To address this issue, the task of top-k sequential pattern mining has been defined, where k is the number of sequential patterns to be found, and is set by the user. In this paper, we present an efficient algorithm for this problem named TKS (Top-K Sequential pattern mining). TKS utilizes a vertical bitmap database representation, a novel data structure named PMAP (Precedence Map) and several efficient strategies to prune the search space. An extensive experimental study on real datasets shows that TKS outperforms TSP, the current state-of-the-art algorithm for top-k sequential pattern mining by more than an order of magnitude in execution time and memory.