Résumé: Recent advancements in information technology have ushered a new wave of systems integrating Internet technology with sensing, wireless communication, and computational resources over existing infrastructures. As a result, myriad complex, non-traditional Cyber-Physical Systems (CPS) have emerged, characterized by interaction among people, physical facilities, and embedded sensors and computers, all generating vast amounts of complex data. Such a case is encountered within a contemporary airport hall setting: passengers roaming, information systems governing various functions, and data being generated and processed by cameras, phones, sensors, and other Internet of Things technology. This setting has considerable potential of contributing to goals entertained by the CPS operators, such as airlines, airport operators/owners, technicians, users, and more. We model the airport setting as an instance of such a complex, data-intensive CPS where multiple actors and data sources interact, and generalize a methodology to support it and other similar systems. Furthermore, this article instantiates the methodology and pipeline for predictive analytics for passenger flow, as a characteristicmanifestation of such systems requiring a tailored approach. Our methodology also draws from DataOps principles, using multi-modal and real-life data to predict the underlying distribution of the passenger flow on a flight-level basis (improving existing day-level predictions), anticipating when and how the passengers enter the airport and move through the check-in and baggage drop-off process. This allows to plan airport resources more efficiently while improving customer experience by avoiding passenger clumping at check-in and security. We demonstrate results obtained over a case from a major international airport in the Netherlands, improving up to 60% upon predictions of daily passenger flow currently in place.

Langue: Anglais