"""
Module holding functions for a quantitative comparison of groupings
"""
from sklearn import metrics
from .spatial import get_groups as get_spatial_groups
[docs]def social_vs_spatial(node_locations, soc_membership, **kwargs):
"""
This method relies on the
:func:`~.sklearn.metrics.homogeneity_completeness_v_measure` method form
the :obj:`.sklearn.metrics` package.
.. seealso::
| V-Measure:
| A conditional entropy-based external cluster evaluation measure
| Andrew Rosenberg and Julia Hirschberg, 2007
| http://acl.ldc.upenn.edu/D/D07/D07-1043.pdf
Parameters
==========
node_locations: dict(int, tuple)
Specify for each node (key) its position (value) in the form of a (x, y)
tuple.
soc_membership: dict
Holds for each node (key) the social group affiliation (value).
**kwargs
All keyword arguments forwarded to :func:`.spatial.get_groups`
Returns
=======
homogeneity, completeness, v-measure: tuple
Homogeneity, completeness and v-measure as computed by the
:func:`~.sklearn.metrics.homogeneity_completeness_v_measure`.
"""
nbr_clusters = len(set(soc_membership.values()))
spatial_membership = get_spatial_groups(
node_locations, nbr_clusters, **kwargs
)
nodes = sorted(soc_membership.keys())
# create lists for both memberships with the same ordering
_soc_membership = [soc_membership[node] for node in nodes]
_spat_membership = [spatial_membership[node] for node in nodes]
h, c, v = metrics.homogeneity_completeness_v_measure(
_soc_membership, _spat_membership
)
return h, c, v
