Mining Frequent Trajectory Using FP-tree in GPS Data ⋆

Pervasiveness of location-acquisition technologies makes it convenient to collect the movement data of moving objects, and the spatial-temporal information contained implicitly in the historical trajectories unveils important knowledge about movement behaviors. This paper presents a novel frequent trajectory mining method using FP-Tree Most existing approaches transform trajectories into sequences of popular region-ids using a statically predefined grid of cells with the same size, and then merge popular cells into larger popular regions. However, due to the size of these popular regions have not been limited, the movements of objects in the region may be lost. And predefined grid may be lack of adaptability. This study defines a Boundary Function to limit the maximum size of the popular regions and selects the size of the grid dynamically by defining a distance threshold d. Then, an improved FP-Tree algorithm is proposed to mine frequent trajectories. The experimental results show our method is efficient