Continuous pattern mining using the FCPGrowth algorithm in trajectory data warehouses

Abstract

This paper presents the FCP-Tree index structure and the new algorithm for continuous pattern mining, called FCPGrowth, for Trajectory Data Warehouses. The FCP-Tree is an aggregate tree which allows storing similar sequences in the same nodes. A characteristic feature of the FCPGrowth algorithm is that it does not require constructing intermediate trees at recursion levels and therefore, it has small memory requirements. In addition, when the initial FCP-Tree is built, input sequences are split on infrequent elements, thereby increasing the compactness of this structure. The FCPGrowth algorithm is much more efficient than our previous algorithm, which is confirmed experimentally in this paper. © 2010 Springer-Verlag.