Road traffic inference and data-driven routing

In this thesis we develop data-driven procedures with the aim to identify individual-specific optimal routes. The approach entails modeling travel times as random variables, such that the travel time distribution of a route provides a full characterization of its travel time properties. While these travel time distributions are often not readily available, we infer them with data generated by users of the road network. After inferring the travel time distributions, we proceed by identifying the route that maximizes an individual's utility function. In doing so, we recognize that statistical uncertainty should be incorporated when determining the optimal route. Each chapter of this thesis outlines a specific data-driven approach, for which we tailor conventional statistical methods to be used with a specific type of data.