Database Reference
In-Depth Information
the probability that a check-in will lead to a new friendship is only 24%. Such
results were confirmed in phone call data, with the influence of friendship on an
individual's mobility about 2
.
5 times greater than the influence of mobility on
creating friendships. Moreover, data also display a strong dependency between
probability of friendship and trajectory similarity, suggesting that there is a
strong presence of social and geographical homophily.
The most interesting aspect of such main findings in the interplay between
sociality and mobility is that they can be used to develop a model of human
mobility dynamics combining periodic daily movement patterns with the social
movement effects coming from the friendship network.
15.3 Conclusions
We have discussed in this chapter how the tools of statistical physics and com-
plexity science have been applied to the study of human mobility, both focusing
on individual movements and considering also the social relations among indi-
viduals. We have observed how, in both cases, general laws can be devised and
empirically validated based on the newly available mobility data, shedding a
new light on the underlying mechanisms behind phenomena that, at first sight,
seem to be governed by chaos.
We conclude with an observation that spontaneously emerges from the cur-
rent trend of research, as presented here: there is an evident push toward the
convergence of network/complexity science and data mining research, a pro-
gressive merge of the two scientific communities that is only beginning today,
but is steadily increasing due to the advantages of combining the complementary
strengths and weaknesses of the two approaches. Why is this merge convenient?
We learned in this chapter that statistical physics and network science are
aimed at discovering the global models of complex social phenomena, by means
of statistical macro-laws governing basic quantities; the ubiquitous presence of
power laws and other long-tailed distributions allows us to witness the behav-
ioral diversity in society at large, such as the huge variability and individual
differences of human movements. On the other hand, data mining is aimed at
discovering local patterns of complex social phenomena, by means of micro-
laws governing behavioral similarity or regularities in subpopulations, such as
the mobility patterns and clusters discussed in Chapters
6
and
7
of this topic.
This dualistic approach is illustrated in Figure
15.8
. In the overall set of indi-
vidual trajectories across a large city we observe a huge diversity: while most
travels are short, a small but significant fragment of travels are extraordinarily
long; therefore, we observe a long-tailed, scale-free distribution of quantities
such as the travel length and the users' radius of gyration. Despite this com-
plexity represented in the data, mobility data mining can automatically discover
travel patterns corresponding to a set of travelers with similar mobility: in such
