String-Based Feature Representation for Trajectory Clustering

Abstract

A trajectory is the spatial trail of a moving object as a function of time. All moving objects such as humans, robots, cloud, taxis, animals, mobile phones generate trajectories. Trajectory clustering is grouping of trajectories that have similar moving patterns, and the formed clusters depend on feature representation, similarity metrics, and clustering algorithm used. In this article, trajectory features are generated after mapping trajectories onto grids, as this smoothens the variations that occur in spatial coordinates. These variations occur due to differences in how GPS points at varying intervals are generated by the device, even when they follow the same path. The main motivation for the article is to devise an algorithm for trajectory clustering that is independent of the variations from GPS devices. A string-based model is used, where trajectories are represented as strings and string-based distance metrics are used to measure the similarity between trajectories. A hierarchical method is applied for clustering and the results are validated using three metrics. An experimental study is conducted and the results show the effectiveness of string-based representation and distance metrics for trajectory clustering.