Understanding Human Mobility During Events in
                     Foursquare

      Cristina Ioana Muntean1 , Franco Maria Nardini1 , and Anastasios Noulas2
                                    1
                                       ISTI-CNR, Pisa, Italy
               2
                   Data Science Institute, Lancaster University, United Kingdom


       Abstract. Social events can generate high influxes of people transitioning var-
       ious locations in a city. They can be considered to have a considerable impact
       on the local economy, whether they are sport events, concerts or festivals. These
       events are capable of generating sudden changes in the activity landscape of a
       city, with the neighborhoods that host events becoming unusually busy and active
       compared to times of regular citizen activity. While event and anomaly detec-
       tion more generally has been a topic of study in recent years, as also has been
       event recommendation for mobile users, progress has been slower towards build-
       ing systems that are able to capture the sudden shift appropriately in this setting.
       In this work we exploit data from the location-based service Foursquare to study
       mobility during events in Chicago, and later expand our study to other cities as
       well. Our aim is to identify what differences emerge in terms of user mobility
       during events versus regular periods of human activity.

    Social events are inherent to urban lifestyle and attract hundreds, even thousands of
people at a designated location, typically for a short period of time. Recent studies show
that the movement landscape in a city alters tremendously during events [1]. There is
a lack of frameworks in the literature of human mobility modeling that are specifically
designed to react in this context, thus adapt and tune mobility models so they remain
effective in event and non-event contexts.
    We perform an analysis and comparison of user mobility patterns during event and
regular (non-event) periods in Chicago extracted from Foursquare LBSN as follows:
-Spatio-tempoal event detection: Initially we apply a method for detecting events
through time by observing anomalies in the popularity of places exploiting statistical
measurements like z-scores in terms of standard deviations from historic popularity.
-Event mobility analysis: Having isolated the scope of events across spatial and tempo-
ral dimensions, we perform a comparative analysis of mobility during event and regular
times with measures for the geographic dispersion of users across space and time.
    Results have shown that while the general trends in human movement, such as pre-
ferring nearby places to distant ones, are preserved, there are significant shifts in the
ways users move during events. They tend to move faster, visit more popular places and
cover longer distances compared to regular times of mobility.
References
1. Petko Georgiev, Anastasios Noulas, and Cecilia Mascolo. Where businesses thrive: Predicting
   the impact of the Olympic Games on local retailers through location-based services data.
   ICWSM, 2014.