Building knowledge graph in the transportation domain

Transportation data are naturally graph-oriented. All our everyday transit decisions are based on oriented paths from A to B in graph topology without us consciously realizing that. So why do we still look at transportation domain data as transaction data, pushing them into relational databases and losing graph analytics opportunities given to us for free? Let's leverage the full potential of transportation data with graph theory, state-of-the-art graph databases, and fast graph algorithms. The paper aims to propose the usage of a modern NoSQL graph-oriented database in the transportation domain. Our goal is to create a transportation knowledge graph: a network of real-world interconnected entities organized in layers. Layers of data from various transportation domains are linked together following the geographical attributes of nodes mapped to the open-street map layer. The ability to see multiple transportation data in one interconnected network is crucial for our next analytical stage. Knowledge graph structure encourages us to analyze the system and solve centrality, pathfinding, or similarity challenges. To demonstrate the application of our approach, we use the PageRank algorithm to explore geographical heat spots based on start and end rides data from student carsharing Uniqway. Uniqway is carsharing made by students of Czech Technical University in Prague (CTU), Czech University of Life Sciences Prague (CZU) and Prague University of Economics and Business (VSE) for university students and employees. Knowledge graph presents a general structure, and future work can go deeper in various fields such as optimization or machine learning.