Trajectory-based Modeling for Fraud Detection and Analytics: Foundation and Design

The trajectory concept is immersed in various socio-technical systems. Certainly, moving object data intended for location-based services is predominant, due to availability of spatiotemporal location data provided by current devices such as mobile devices, GPS devices, RFID and various sensors. This work examines the use of our trajectories meta-model given in previous works to detect and monitor frauds, from the point of view of trajectories semantic aspects. Modeling of frauds by trajectories gives meaning to the sequence of events that made it possible to detect them and to recognize the situations of their occurrence. The semantic dimension qualifies a fraud trajectory as a story that has both meaning and intention, described by episodes of semantic events related to the creation of the fraud, to its dynamic and to its ending. Fraud analytics will benefit from trajectory analytics for both big data ecosystems and spatiotemporal data mining algorithms and practices. Finally, we propose the architecture and structure of the software components that make up the proposed software system. The software design is based on the reactive manifesto and federates graph-oriented NoSQL databases and graph-based data science processing.