Data from: Prediction limits of mobile phone activity modelling

Thanks to their widespread usage, mobile devices have become one of the
main sensors of human behaviour and digital traces left behind can be used
as a proxy to study urban environments. Exploring the nature of the
spatio-temporal patterns of mobile phone activity could thus be a crucial
step towards understanding the full spectrum of human activities. Using 10
months of mobile phone records from Greater London resolved in both space
and time, we investigate the regularity of human telecommunication
activity on urban scales. We evaluate several options for decomposing
activity timelines into typical and residual patterns, accounting for the
strong periodic and seasonal components. We carry out our analysis on
various spatial scales, showing that regularity increases as we look at
aggregated activity in larger spatial units with more activity in them. We
examine the statistical properties of the residuals and show that it can
be explained by noise and specific outliers. Also, we look at sources of
deviations from the general trends, which we find to be explainable based
on knowledge of the city structure and places of attractions. We show
examples how some of the outliers can be related to external factors such
as specific social events.