Geoscience Reference
In-Depth Information
of the actual typical usage for them to take proactive decisions upon. Diffusing the
knowledge of the signature patterns could also generate new business ideas in the
cities, by just allowing retailers to optimize scheduling based on the likelihood of
people being in the vicinity of their stores.
We presented in Sect.
15.5
a clustering analysis process allowing automatic
detection of locations with similar signatures within a city. Our findings showed that
we could detect in each city a core business center, at least one pure residential area,
one mixed area with a strong commercial component, and another mixed area with
a strong residential component. In addition, we were able to detect other clusters of
specific nature (commuting hub areas in London or the JFK airport in New York).
The methods used could be easily pushed to generalize these findings. We stressed
that our approach makes it possible to build an automatic, quick, and relatively
cheap way of preparing maps that could complement and enrich classical land use
maps based on surveys. Indeed, communication traffic data tell us about the actual
dynamic behavior of people at each location, while land use maps rather tell us
about the average type of behavior you can expect based on the urbanization levels
and the type of buildings, shops, or infrastructures present at each location.
Our final finding was obtained by applying our clustering the procedure on the
locations of all three cities at once. Quite surprisingly, we found that the core
business centers of London, New York, and Hong Kong were gathered into a single
cluster, which proves their high degree of similitude. On the other hand, the city has
residential locations whose signatures are well distinct. To answer a question raised
at the beginning of this chapter, it seems like globalization shapes the economical
and political activity in the large cities' financial and decisional core centers, while
individual activity patterns are still defined mostly by local factors.
On a broader note, as the digital databases are growing and our computational
methods are improving, it may be tempting to multiply automatic procedures to
generate lists of insights based on mobile traffic data. However, we argue that
knowledge expertise is more than ever needed to understand, interpret, and critic
these results.
Though this chapter presented new similarity measurements and links between
three major cities, it also opens up the question of the universality of our findings.
Is the common beat detected in the core areas a universal pattern common to
all cities, or is it only peculiar to “occidental” developed cities? Is there any
natural classification of the world's cities based on a similarity between peripheral
residential signatures? It would be most interesting to study these questions by
enlarging our datasets to include major cities from both developed and developing
countries such as Paris, Mexico City, Shanghai, Rio de Janeiro, Sao Paulo, Lagos,
or Mumbai. The challenge is now to gather a collection of mobile traffic data from
all these cities before starting to build a yet-to-define comparative science of cities
based on them.
Acknowledgements
We thank Ericsson for providing datasets for this research and especially
Dwight Witherspoon for the organizational support to the project. We also thank Christine Maynié-
François for the stimulating discussions and thorough proofreading. We would further like to thank
the National Science Foundation, the AT&T Foundation, the MIT SMART program, the Center
Search WWH ::
Custom Search