Interpreting Traffic Congestion Using Fundamental Diagrams and Probabilistic Graphical Modeling

Traffic congestion is a major economic, environmental and social issue that affects cities throughout the world. This research explains the complex associations of traffic flow based in an empirical-theoretical framework using real-world datasets. We propose a data fusion method to infer well-defined microscopic fundamental diagrams in dense urban areas making use of inductive loop detectors and taxi trajectory data. We also present a semi-naive Bayesian modeling approach to extract causality knowledge built on previous discriminated congestion in different road segments. A realistic empirical evaluation allows us to identify and quantify causalities between congestion and diverse confounding variables (e.g. meteorological conditions). Our aim is to contribute to efficient traffic flow by uncovering the tangled traffic congestion in an urban geographical area.