Mining Disjoint Sequential Pattern Pairs from Tourist Trajectory Data

Abstract

Route mining from trajectory databases, or trajectory data mining, has become an important and valuable task since the popularization of GPS devices. Sequential pattern mining based approaches are well applied to trajectory data mining, while they often suffer the problem of high redundancy and low comprehensibility. In this paper, we solve this problem by proposing a novel approach of disjoint sequential pattern pair mining, which takes on a new perspective to this problem by focusing on extracting extra valuable information, i.e., hyper patterns from the “redundant” patterns instead of just removing them. We conduct experiments on a real tourist trajectory database as well as an artificial one. We show the practical applicability of our approach and the effectiveness and efficiency of our mining algorithm by analyzing the mining results.