def test_community_analysis_single_channel_cutoff_20(self):
log_data = reader.linux_input(self.log_data_dir, ["#kubuntu-devel"],
self.start_date, self.end_date)
expected_result = util.load_from_disk(
self.current_directory +
'/data/output/community_analysis_single_channel_cutoff_20')
nicks, nick_same_list = nickTracker.nick_tracker(log_data)
default_cutoff = config.THRESHOLD_MESSAGE_NUMBER_GRAPH
config.THRESHOLD_MESSAGE_NUMBER_GRAPH = 20
message_number_graph = network.message_number_graph(
log_data, nicks, nick_same_list, False)
saver.save_net_nx_graph(
message_number_graph, self.current_directory,
"message-exchange-" + self.start_date + "-cutoff-" +
str(config.THRESHOLD_MESSAGE_NUMBER_GRAPH))
expected_output = community.infomap_igraph(
ig_graph=None,
net_file_location=self.current_directory + "/message-exchange-" +
self.start_date + "-cutoff-" +
str(config.THRESHOLD_MESSAGE_NUMBER_GRAPH) + '.net')
os.remove(self.current_directory + "/message-exchange-" +
self.start_date + "-cutoff-" +
str(config.THRESHOLD_MESSAGE_NUMBER_GRAPH) + '.net')
config.THRESHOLD_MESSAGE_NUMBER_GRAPH = default_cutoff
self.assertTrue(expected_result[0].isomorphic(expected_output[0]))
self.assertEqual(
compare_communities(expected_result[1], expected_output[1]), 0)
def test_infomap_igraph(self):
message_graph, message_comm = community.infomap_igraph(ig_graph=None, net_file_location= self.current_directory + '/data/message_number_graph.net')
expected_result = util.load_from_disk(self.current_directory + '/data/community')
dis = clustering.compare_communities(message_comm, expected_result) #calculate distance between two communities
self.assertEqual(dis, 0.0)
def test_community_analysis_multi_channel(self):
log_data = reader.linux_input(self.log_data_dir, ["ALL"],
self.start_date, self.end_date)
expected_result = util.load_from_disk(
self.current_directory +
'/data/output/community_analysis_multi_channel')
nicks, nick_same_list, channels_for_user, nick_channel_dict, nicks_hash, channels_hash = nickTracker.nick_tracker(
log_data, True)
dict_out, graph = network.channel_user_presence_graph_and_csv(
nicks, nick_same_list, channels_for_user, nick_channel_dict,
nicks_hash, channels_hash)
presence_type = ["CC", "UU", "CU"]
expected_output = {ptype: {} for ptype in presence_type}
for ptype in presence_type:
saver.save_net_nx_graph(dict_out[ptype]["graph"],
self.current_directory, "adj" + ptype)
saver.save_net_nx_graph(dict_out[ptype]["reducedGraph"],
self.current_directory, "radj" + ptype)
expected_output[ptype]['adj'] = community.infomap_igraph(
ig_graph=None,
net_file_location=self.current_directory + '/adj' + ptype +
'.net')
expected_output[ptype]['radj'] = community.infomap_igraph(
ig_graph=None,
net_file_location=self.current_directory + '/radj' + ptype +
'.net')
os.remove(self.current_directory + '/adj' + ptype + '.net')
os.remove(self.current_directory + '/radj' + ptype + '.net')
self.assertTrue(expected_result[ptype]['adj'][0].isomorphic(
expected_output[ptype]['adj'][0]))
self.assertEqual(
compare_communities(expected_output[ptype]['adj'][1],
expected_result[ptype]['adj'][1]), 0)
self.assertTrue(expected_result[ptype]['radj'][0].isomorphic(
expected_output[ptype]['radj'][0]))
self.assertEqual(
compare_communities(expected_output[ptype]['radj'][1],
expected_result[ptype]['radj'][1]), 0)
#%%
argv = ['a', '/Users/hamishgibbs/Documents/Covid-19/unequal_mobility_uk/output/spi_m/infomap/cluster_full.csv']
im = pd.read_csv(argv[1])
im['quadkey'] = [f'{n:012}' for n in im.quadkey.values]
#%%
im = im.groupby('date')
im = [im.get_group(x) for x in im.groups]
#%%
i0 = im[0][['quadkey', 'cluster']].rename(columns = {'cluster':'cluster_0'})
i1 = im[1][['quadkey', 'cluster']].rename(columns = {'cluster':'cluster_1'})
#%%
ic = pd.merge(i0, i1).dropna(axis = 0, how = 'any')
compare_communities(ic['cluster_0'].tolist(), ic['cluster_1'].tolist(), method = 'nmi')
#%%
nmi = []
for i, df in enumerate(im):
if i > 0:
i0 = im[i - 1][['quadkey', 'cluster']].rename(columns = {'cluster':'cluster_0'})
i1 = im[i][['quadkey', 'cluster']].rename(columns = {'cluster':'cluster_1'})
ic = pd.merge(i0, i1).dropna(axis = 0, how = 'any')
nmi.append(compare_communities(ic['cluster_0'].tolist(), ic['cluster_1'].tolist(), method = 'nmi'))
#%%
import matplotlib.pyplot as plt
# first parameter group
for din in range(9, 18):
for _ in range(100):
file = DIR_FMT.format(64) + GML_FMT.format(din, _)
g = ig.Graph.Read_GML(file)
ground_truth = [int(v['value']) for v in g.vs]
for __ in range(100):
p1 = g.community_fastgreedy().as_clustering()
p2 = g.community_infomap()
p3 = g.community_label_propagation()
p4 = g.community_multilevel()
p5 = g.community_walktrap().as_clustering()
Fastgreedy_performance[0].append(
compare_communities(ground_truth, p1, method='nmi'))
Infomap_performance[0].append(
compare_communities(ground_truth, p2, method='nmi'))
LPA_performance[0].append(
compare_communities(ground_truth, p3, method='nmi'))
Louvein_performance[0].append(
compare_communities(ground_truth, p4, method='nmi'))
Walktrap_performance[0].append(
compare_communities(ground_truth, p5, method='nmi'))
g = nx.read_gml(file, label='id')
partitions = FPPM_repeat(g)
for p in partitions:
FPPM_performance[0].append(
compare_communities(ground_truth, p, method='nmi'))
for idx, N in enumerate(SIZES):
for Mu in range(1, 10):
for _ in range(100):
file = DIR_FMT.format(N) + GML_FMT.format(Mu, _)
g = ig.Graph.Read_GML(file)
ground_truth = [int(v['value']) for v in g.vs]
for __ in range(100):
p1 = g.community_fastgreedy().as_clustering()
p2 = g.community_infomap()
p3 = g.community_label_propagation()
p4 = g.community_multilevel()
p5 = g.community_walktrap().as_clustering()
Fastgreedy_performance[idx].append(compare_communities(ground_truth, p1, method='nmi'))
Infomap_performance[idx].append(compare_communities(ground_truth, p2, method='nmi'))
LPA_performance[idx].append(compare_communities(ground_truth, p3, method='nmi'))
Louvein_performance[idx].append(compare_communities(ground_truth, p4, method='nmi'))
Walktrap_performance[idx].append(compare_communities(ground_truth, p5, method='nmi'))
g = nx.read_gml(file, label='id')
partitions = FPPM_repeat(g)
for p in partitions:
FPPM_performance[idx].append(compare_communities(ground_truth, p, method='nmi'))
with open('LFR_benchmarks/Fastgreedy_performance.pickle', 'wb') as f:
pickle.dump(Fastgreedy_performance, f)
with open('LFR_benchmarks/Infomap_performance.pickle', 'wb') as f:
pickle.dump(Infomap_performance, f)
with open('LFR_benchmarks/LPA_performance.pickle', 'wb') as f: