|Institution:||Colorado School of Mines|
|Full text PDF:||http://hdl.handle.net/11124/170018|
Commuting networks describe the flows of individuals from one location to another. These networks are present in many different application scenarios, including traffic modeling, infrastructure planning, and epidemic simulation. Traditionally, commuting networks are created using data from costly and outdated surveys. This dissertation shows how individual's location information can be data mined from social media communication and be used to build commuting networks. Some of the problems discussed in this dissertation include the quality aspects of location information obtained from social media and the lack of representation of social media users in the overall general population. Two models for commuting networks, the gravity model and the radiation model, are described and evaluated. This dissertation also presents GeoDigger, a tool that can be used to help researchers collect location information from Twitter, one of the most popular online social networks. GeoDigger can exclude non-human social activity based on a machine learning technique adapted to work with imbalanced data. Advisors/Committee Members: Camp, Tracy (advisor), Munjal, Aarti (advisor), Navidi, William Cyrus (committee member), Mehta, Dinesh P. (committee member), Aschenbruck, Nils (committee member), Perrone, Luiz Felipe (committee member).