      Ahmed Kharrat 1 , Iulian Sandu Popa 1 Karine Zeitouni 1 , Sami Faiz 2 , 1 PRiSM Laboratory, University of Versailles 45, avenue des Etats-Unis - 78035 Versailles, France 2 LTSIRS, Ecole nationale d’ingénieurs de Tunis B.P. 37 – 1002 Tunis-Belvédère, Tunisie Abstract. Spatial data mining is an active topic in spatial databases. This paper proposes a new clustering method for moving object trajectories da- tabases. It applies specifically to trajectories that only lie on a predefined network. The proposed algorithm (NETSCAN) is inspired from the well- known density based algorithms. However, it takes advantage of the net- work constraint to estimate the object density. Indeed, NETSCAN first computes dense paths in the network based on the moving object count, then, it clusters the sub-trajectories which are similar to the dense paths. The user can adjust the clustering result by setting a density threshold for the dense paths, and a similarity threshold within the clusters. This paper describes the proposed method. An implementation is reported, along with experimental results that show the effectiveness of our approach and the flexibility allowed by the user parameters. Keywords: Spatial data mining, clustering algorithm, similarity measure, moving objects database, road traffic analysis.   Trajectory database management is a relatively new topic of database re- search, which has emerged due to the profusion of mobile devices and po- sitioning technologies like GPS or recently the RFID (Radio Frequency Identification). Trajectory similarity search forms an important class of