The ranking also allows us to compare and validate behavioral
patterns of different groups. For example, we
would like to compare how close the ranking results are
between Flickr, Panoramio and TripAdvisor communities.
Many approaches to achieve this exist: we could use parallel
coordinates, statistical measures or tabular comparison.
We decided to use Andrews Plot, which is proven
to be a good graphic method for detecting outliers [1] in
multivariate data. The plotting is performed by converting
n-dimensional data into a function in a two-dimensional
space of the form:
fx(t) = x1
p
2 + x2sin(t) + x3cos(t) + x4sin(2t) + : : : ;
(2)
where