Towards a Comparative Science of Cities: Using Mobile Traffic Records in New York, London, and Hong Kong - Computational Approaches for Urban Environments - page 373

￿ Similarly, the previously identified “commercial” areas are roughly gathered into

a single cluster - here orange. The correspondence between this orange cluster
and those found in city-independent analyses appears evident for New York, but
less obvious in London and Hong Kong (see Figs. 15.6 a- 15.8 a).

￿ The other clusters mostly correspond to the previously identified residential

areas. Surprisingly, these clusters are almost completely concentrated in one city:
the blue cluster is specific to London, the green cluster is specific to New York,
and the pink and pastel clusters are specific to Hong Kong.

Concerning the residential area, the transversal clustering analysis emphasizes
the differences due to local cultural, technological, and economical factors identified
in Sect. 15.4.2 , e.g., the evening peaks of SMS in New York or the evening peaks of
data transfer in Hong Kong. The very strong and somewhat surprising result here is
the fact that the studied cities have core business centers that share a similar pattern
despite those local factors.

15.6 Discussion

Our research findings demonstrate that a general understanding of the mobile
network signatures can help us to look at cities with a renewed perspective.
We saw in Sect. 15.3 how time-aggregated maps of mobile traffic inhomo-
geneities could capture spatial patterns revealing locations where people are in
general most active. This approach allows to track the location where people spend
most of their time and is complementary to more traditional census data recording
where people live or work. Doing a similar analysis on specific periods of time,
one can expect a rather good correlation between mobile phone activity and job
density, especially during working hours, and a relatively less good correlation
with residential density during working hours that would increase when people
are typically at home (late in the evenings, early in the mornings, or during the
weekends).
In addition, these maps are more up to date than those based on census polling,
relatively cheaper to obtain and dynamic. From a research point of view, one could
also imagine to use insights gained from such representation (like the Gini indices
measuring spatial inhomogeneities) to enrich current taxonomies of cities.
In Sect. 15.4 , we defined typical week patterns or signatures to characterize
the activities' dynamics at a city or local scale. By comparing city signatures, we
highlighted specific influencing factors (mobile traffic plan policy, technological,
economical, and cultural factors) shaping those dynamics in Greater London, New
York, and Hong Kong. Building on the example of a few selected locations within
London, we showed how the signatures could reveal the nature (either financial,
commercial, recreational, residential, or commuting hub) of the concerned areas. In
general, the insights gained from the study of the typical week signatures could be
used to optimize the overall network performance by informing the mobile operators