The Geography of Leisure Preferences from Mobility Data

 

DANIELE QUERCIA

MIT SENSEABLE CITY LAB/UNIVERSITY OF CAMBRIDGE

 

Predicting leisure preferences of people in specific locations makes it possible for marketing agencies to deploy advertising campaigns in targeted locations, and for traffic planners to predict geographical areas likely to originate traffic for specific social events. Traditionally, to determine how user preferences are geographically distributed, marketing agencies run electronic marketing programs, and traffic planners administer surveys. Unfortunately, the activities of electronic marketing and survey administration may suffer from sampling-bias, are expensive, and quickly become out-of-date. Fortunately, researchers have recently mined web 2.0 sites to extract user preferences in specific locations. They have extracted preferences upon digital information generated by individuals. By contrast, we propose to mine information implicitly generated by people while they carry their mobile phones. The idea is that, by mining mobility traces, we are able to infer preferences for social events in a given location. We process 130 millions of anonymous location estimations - latitude and longitude - from roughly 1 million mobile phones in greater Boston (an area of 15km^2). We crawl the "Boston Globe Calendar" website to extract social events. We show that, by analyzing mobility traces and social event data, we are able to label locations with representative social events.