def community_evaluation_metrics(cdlib_coms):
uG = cdlib_coms.graph
# https://cdlib.readthedocs.io/en/latest/reference/evaluation.html
embeddedness = evaluation.avg_embeddedness(uG, cdlib_coms, summary=False)
average_internal_degree = evaluation.average_internal_degree(uG,
cdlib_coms,
summary=False)
conductance = evaluation.conductance(uG, cdlib_coms, summary=False)
transitivity = evaluation.avg_transitivity(uG, cdlib_coms, summary=False)
cut_ratio = evaluation.cut_ratio(uG, cdlib_coms, summary=False)
expansion = evaluation.expansion(uG, cdlib_coms, summary=False)
edges_inside = evaluation.edges_inside(uG, cdlib_coms, summary=False)
fraction_over_median_degree = evaluation.fraction_over_median_degree(
uG, cdlib_coms, summary=False)
hub_dominance = evaluation.hub_dominance(uG, cdlib_coms, summary=False)
internal_edge_density = evaluation.internal_edge_density(uG,
cdlib_coms,
summary=False)
max_odf = evaluation.max_odf(uG, cdlib_coms, summary=False)
avg_odf = evaluation.avg_odf(uG, cdlib_coms, summary=False)
flake_odf = evaluation.flake_odf(uG, cdlib_coms, summary=False)
size = evaluation.size(uG, cdlib_coms, summary=False)
triangle_participation_ratio = evaluation.triangle_participation_ratio(
uG, cdlib_coms, summary=False)
eval_dict = {
'embeddedness': embeddedness,
'average_internal_degree': average_internal_degree,
'conductance': conductance,
'transitivity': transitivity,
'cut_ratio': cut_ratio,
'expansion': expansion,
'edges_inside': edges_inside,
'fraction_over_median_degree': fraction_over_median_degree,
'hub_dominance': hub_dominance,
'internal_edge_density': internal_edge_density,
'max_odf': max_odf,
'avg_odf': avg_odf,
'flake_odf': flake_odf,
'size': size,
'triangle_participation_ratio': triangle_participation_ratio
}
com_eval_df = pd.DataFrame(eval_dict)\
.reset_index()\
.rename({'index':'community_id'}, axis = 1)
com_eval_df['community_id'] = com_eval_df['community_id'] + 1
return com_eval_df
def flake_odf(self, **kwargs):
"""
Fraction of nodes in S that have fewer edges pointing inside than to the outside of the algorithms.
.. math:: f(S) = \\frac{| \{ u:u \in S,| \{(u,v) \in E: v \in S \}| < d(u)/2 \}|}{n_S}
where :math:`E` is the graph edge set, :math:`v` is a node in :math:`S`, :math:`d(u)` is the degree of :math:`u` and :math:`n_S` is the set of algorithms nodes.
:param summary: (optional, default True) if **True**, an overall summary is returned for the partition (min, max, avg, std); if **False** a list of community-wise score
:return: a FitnessResult object/a list of community-wise score
:Example:
>>> from cdlib.algorithms import louvain
>>> g = nx.karate_club_graph()
>>> communities = louvain(g)
>>> mod = communities.flake_odf()
"""
if self.__check_graph():
return evaluation.flake_odf(self.graph, self, **kwargs)
else:
raise ValueError("Graph instance not specified")