Zombies Walk Among Us: Cross-platform data mining for event monitoring

Data mining across multiple social websites
can reveal valuable factual information for both monitoring and
reconstructing events. Crowdsourcing can be used on these sites
to monitor ‘flash mob’ group behaviours, loosely or formally
planned activities, such as ‘zombie walks’. In certain contexts,
these walks are ‘marketed’ to social site users in order to
promote charitable or social engagements; in others, analysis
suggests that participation is a form of political engagement. We
use multi-platform information extraction to build an atlas of
geographic and demographic events, which leads us to compelling,
yet imperfect, understanding of why events are successful and
regionally relevant factors that encourage people to pick up
behaviours or participate in movements. More generally, this case
study confirms to us that that ’big-data’-led analysis of this kind is
reasonably straightforward and rewarding, as well as providing a
useful basis for Bayesian reasoning: the use of available evidence
to evaluate propositions about this type of event.