Review and comparison of similarity measures and community detection algorithms for clustering of network constrained trajectories

Noudéhouénou Lionel Jaderne Houssou (1) , Jean-Loup Guillaume , Armelle Prigent (1)

1 L3I - Laboratoire Informatique, Image et Interaction - EA 2118

Trajectory analysis is a study field that is experiencing a renewed interest mainly due to the increase and availability of datasets, generated by users either in the real world by the means of sensors such as GPS, or in virtual environments thanks to the digital footprint they leave when visiting a commercial website for example. In this paper we present the process of converting raw trajectories into network-constrained ones and we review similarity measures for trajectories as well as how they can be used to represent trajectories as graphs. We also describe some approaches to cluster graph of trajectories. Then we assess the trajectories similarity measures and the algorithms for graph clustering. In particular we show that some similarity measures are inadequate for clustering graph of trajectories since they are not dis-criminative enough. RÉSUMÉ.

Mots clés: complex networks; network-constrained trajectories; trajectory similarity; trajectory clustering

Mots clés: graphes de terrain; trajectoires contraintes par un réseau; similarité de trajectoires; clustering de trajectoires

Domaines: Informatique [cs]

Noudéhouénou Lionel Jaderne Houssou, Jean-Loup Guillaume, Armelle Prigent. Review and comparison of similarity measures and community detection algorithms for clustering of network constrained trajectories. MARAMI, Oct 2017, La Rochelle, France. ⟨hal-02363974⟩