This presentation is devoted to a recent TAMALE project in which we show how Machine Learning can be applied to a practical problem in fisheries. High seas fish stocks are managed by regional fisheries management organizations (RFMOs) composed of members from different fishing nations. These regional regimes are responsible for the conservation and protection of fish stocks. RFMOs set and allocate quotas for the fish stocks under their management within the boundaries set out in their conventions. They are also responsible for enforce their quotas through control, monitoring and surveillance activities. One aspect of monitoring involves determination if a vessel is fishing or not during a given time interval. The vessel monitoring system (VMS) is a satellite-based monitoring system which at regular intervals provides data to the fisheries authorities on the location, course and speed of vessels. At present, VMS data do not indicate whether a vessel is fishing when its position is reported. Therefore, the use of VMS data to estimate fishing effort depends on accurate differentiation between fishing and nonfishing activity. This paper introduces the conceptual model of trajectories for users of VMS and also conducts the first field study of trajectories using supervised learning because. Until now, for lack information inherent in the moving object during the development of their journey, the trajectories are analyzed by clustering methods based on unsupervised learning.