Pinheiro, Flávio Luís PortasNeto, Miguel de Castro Simões FerreiraPanchot, Alex Anthony2021-09-032021-09-032021-07-28http://hdl.handle.net/10362/123740Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced AnalyticsWhile some analytical solutions approach a problem with existing data, agent-based simulations give the ability to approach the problem with data from the ”future”. The problem tackled in this work is the prediction of specific traffic patterns (namely traffic jams) for the city of Lisbon. Specifically, the question was to see if we can model the change in traffic after a road closure, such as for construction. This question is an interesting one as it explores a logistical problem that every driver and city planner has faced. Additionally while on the surface it seems focused on just traffic and city streets, general problems of path finding and global optimization are also explored. While there exist other software to simulate street traffic, none are very simple and require every location to be manually created for that particular software. The proposed solution is twofold. The first is to design a reproducible method for creating the agents from either existing data or from a set of rules that can be used to create agents where there is no existing data. The second is to design a simulation algorithm that can run off of an existing map, in this case open-source OpenStreetMap, such that the map can instantly be swapped with any other with no changes to the code required. By designing a simulation like this, the ability to simultaneously compare multiple modes of transportation can also be realized. The final results show that this project is successful as a way of seeing at least general trends as a result of a change in the road network (a road closure). While more work is needed to add multiple modes of transportation, the ability for the simulation to run on a different road network allows it to be easily extended in this area as well as other graph networks or flow simulations. While not perfect, simulations can be a significantly cheaper and safer option for analyzing changes to a network than actually making that change and analyzing what happened after the fact.engAgent Based SimulationsHuman Mobility Simulationsmaster thesis202762807