Managing data of sensor-equipped transportation networks using graph databases

Abstract. In this paper, we are concerned with data pertinent to transportation networks, which model situations in which objects move along a graph-like structure. We assume that these networks are equipped with sensors that monitor the network and the objects moving along it. These sensors produce time series data, resulting in sensor networks. Examples are river, road, and electricity networks. Geographical information systems are used to gather, store, and analyse data, and we focus on these tasks in the context of data emerging from transportation networks equipped with sensors. While tailored solutions exist for many contexts, they are limited for sensor-equipped networks at this moment. We view time series data as temporal properties of the network and approach the problem from the viewpoint of property graphs. In this paper, we adapt and extend the theory of the existing property graph databases to model spatial networks, where nodes and edges can contain temporal properties that are time series data originating from the sensors. We propose a language for querying these property graphs with time series, in which time series and measurement patterns may be combined with graph patterns to describe, retrieve, and analyse real-life situations. We demonstrate the model and language in practice by implementing both in Neo4j and explore questions hydrology researchers pose in the context of the Internet of Water, including salinity analysis in the Yser river basin.