def read_nets_by_years(path):
filenames = glob.glob(path)
filenames = sorted(filenames)
dates = []
nets = []
for filename in filenames:
net = xnet.xnet2igraph(filename)
net.vs['political_party'] = [
filter_pp_name(p) for p in net.vs['political_party']
]
nets.append(net.components().giant())
base = filename.split('dep')[1].split('_')
# date = float(filename.split('dep')[1].split('_')[0])
date = float(filename.split('_')[2].split('.')[0])
dates.append(date)
dates = np.asarray(dates)
nets = np.asarray(nets)
sorted_idxs = np.argsort(dates)
dates = dates[sorted_idxs]
nets = nets[sorted_idxs]
return dates, nets
def xnet_input_to_wordcloud(input_file, output_file):
graph = xn.xnet2igraph(input_file)
wc = WordCloud(background_color="white",
max_words=2000,
scale=10,
contour_width=3,
contour_color='white')
wc.generate("\n".join(graph.vs["paper_abstract"]))
wc.to_file(output_file)
def calculate_dist(filenames):
for filename in filenames:
# print(filename)
net = xnet.xnet2igraph(filename)
weights = net.es['weight']
weights = [math.sqrt(2 * (1 - w)) for w in weights]
if len(weights) > 0:
net.es['distance'] = weights
xnet.igraph2xnet(net, filename[:-5] + "_dist.xnet")
else:
print('error', filename)
def xnet_input_to_communities_wordcloud(input_file,
output_file,
minYear=minYear,
minKCore=minKCore):
graph = xn.xnet2igraph(input_file)
verticesToDelete = np.where(
np.logical_or(
np.array(graph.vs["year"]) < minYear,
np.array(graph.vs["KCore"]) < minKCore))[0]
graph.delete_vertices(verticesToDelete)
graph = graph.clusters(mode="WEAK").giant()
communities = graph.vs["Community"]
sortedCommunities = sortByFrequency(communities)[0:maxCommunities]
fig = plt.figure(figsize=(20, 5 * math.ceil(len(sortedCommunities) / 2)))
allAbstracts = "\n".join(graph.vs["paper_abstract"])
allFrequencies = WordCloud(
max_words=maxAllWords).process_text(allAbstracts)
amask = np.zeros((500, 1000), dtype='B')
amask[:10, :] = 255
amask[-10:, :] = 255
amask[:, :10] = 255
amask[:, -10:] = 255
for index, community in enumerate(sortedCommunities):
communityColor = (_styleColors[index]
if index < len(_styleColors) else "#aaaaaa")
abstracts = "\n".join([
vertex["paper_abstract"] for vertex in graph.vs
if vertex["Community"] == community
])
plt.subplot(math.ceil(len(sortedCommunities) / 2), 2, index + 1)
wc = WordCloud(background_color="white",
max_words=maxInternalWords,
width=1000,
height=500,
mask=amask,
contour_width=10,
contour_color=communityColor,
random_state=3,
color_func=generateColorFunction(communityColor))
inCommunityFrequency = wc.process_text(abstracts)
relativeFrequencies = {
key: frequency / math.log(allFrequencies[key] + 1)
for key, frequency in inCommunityFrequency.items()
if key in allFrequencies
}
wc.generate_from_frequencies(relativeFrequencies)
plt.imshow(wc, interpolation='bilinear')
plt.axis("off")
plt.tight_layout()
plt.savefig(output_file)
plt.close(fig)
def get_areas_all_years():
f = 'data/subj_areas/nets/all.xnet'
g = xnet.xnet2igraph(f)
print(g.vs.attributes())
i = 0
for c in g.community_multilevel(weights='weight'):
for idx in c:
#g.vs[idx]['comm'] = i
print(g.vs[idx]['name'], end=' ')
i += 1
print()
return g
def get_areas_by_year():
files = glob.glob('data/subj_areas/nets/all_with_comm_*.xnet')
gs = []
for file in files:
gs.append(xnet.xnet2igraph(file))
year = 2008
comms = set(gs[0].vs['comm'])
map_to_all = [[(year, c)] for c in comms]
for i in range(len(gs) - 3):
g1 = gs[i]
g2 = gs[i + 1]
comms1 = defaultdict(lambda: [])
comms2 = defaultdict(lambda: [])
for v in g1.vs:
comms1[v['comm']].append(v['name'])
for v in g2.vs:
comms2[v['comm']].append(v['name'])
if len(comms1) <= 0 or len(comms2) <= 0:
continue
for c1, vtxs1 in comms1.items():
sims = []
for c2, vtxs2 in comms2.items():
sim = jaccard(vtxs1, vtxs2)
sims.append((sim, c2))
s_max = max(sims)
if s_max[0] < 0.2:
continue
print(sims)
stop = False
N = len(map_to_all)
for j in range(N):
m = map_to_all[j]
if m[-1][1] == c1 and year == m[-1][0]:
m.append((year + 1, s_max[1]))
stop = True
break
if not stop:
map_to_all.append([(year, c1), (year + 1, s_max[1])])
year += 1
return map_to_all
def xnet_input_to_wordcloud(input_file, output_file):
graph = xn.xnet2igraph(input_file)
wc = WordCloud(background_color="white",
max_words=2000,
scale=10,
contour_width=3,
contour_color='white')
if ("paper_abstract" not in graph.vertex_attributes()):
textData = graph.vs["original_title"]
else:
textData = [
"%s. %s" % (titleAndAbstract) for titleAndAbstract in zip(
graph.vs["original_title"], graph.vs["paper_abstract"])
]
wc.generate("\n".join(textData))
wc.to_file(output_file)
def main():
file_path = '../data/wosAPSWithPACS_WithMAG_raw.xnet'
g = xnet2igraph(file_path)
data = {
'abstract': g.vs['Title and Abstract'],
'year': g.vs['Year Published'],
'language': g.vs['Language'],
'adj_list': g.get_adjlist()
}
df = pd.DataFrame.from_dict(data)
df['adj_list'] = df['adj_list'].astype(str)
df.to_feather('data/all_data.feather')
def plot_(gauss, bins, map_to_all):
fig, axs = plt.subplots(len(map_to_all),
1,
sharex=True,
sharey=True,
figsize=(12, 3 * len(map_to_all)))
for i, group in enumerate(map_to_all):
if len(group) == 1:
continue
print(i, group)
incr, decr = [], []
for year, c in group:
dois = []
f = glob.glob('data/subj_areas/nets/all_with_comm_%s.xnet' %
year)[0]
# print(f)
g = xnet.xnet2igraph(f)
words = set(g.vs.select(comm_eq=c)['name'])
for doi, paper in complete_data.items():
for w in paper['infos']['subj_areas']:
if w in words:
dois.append(doi)
break
dois = set(dois)
incr_temp, decr_temp = incr_decr(dois, year, year + 1)
incr += incr_temp
decr += decr_temp
range0 = (2008, 2020)
hist0, bins_edges0 = np.histogram(incr, bins=bins, range=range0)
hist1, bins_edges1 = np.histogram(decr, bins=bins, range=range0)
# plt.figure(figsize=(12,3))
y0 = np.convolve(hist0, gauss, mode='same')
y1 = np.convolve(hist1, gauss, mode='same')
x = np.arange(bins_edges0[0], bins_edges1[-1],
(bins_edges1[-1] - bins_edges1[0]) / len(y0))
axs[i].set_title("%d - %d" % (group[0][0], group[-1][0]))
axs[i].bar(x - 0.05, y0, width=0.05, label='incr')
axs[i].bar(x, y1, width=0.05, label='decr')
axs[i].legend()
plt.tight_layout()
plt.savefig('hist_areas.pdf')
def xnet_input_to_figure(input_file,
output_file,
minYear=minYear,
minKCore=minKCore):
originalGraph = xn.xnet2igraph(input_file)
graph = originalGraph.clusters(mode="WEAK").giant()
verticesToDelete = np.where(
np.logical_or(
np.array(graph.vs["year"]) < minYear,
np.array(graph.vs["KCore"]) < minKCore))[0]
graph.delete_vertices(verticesToDelete)
graph = graph.clusters(mode="WEAK").giant()
indegree = graph.indegree()
maxIndegree = max(indegree)
graph.vs["vertex_size"] = [x / maxIndegree * 10 + 4 for x in indegree]
colormap = plt.get_cmap("plasma")
if ("Community" not in graph.vertex_attributes()):
graph.vs["color"] = [
convertColorToRGBAString(*colormap(math.log(value + 1)))
for value in indegree
]
else:
communities = graph.vs["Community"]
sortedCommunities = sortByFrequency(communities)
communityToColor = {
community:
(_styleColors[index] if index < len(_styleColors) else "#aaaaaa")
for index, community in enumerate(sortedCommunities)
}
graph.vs["color"] = [
communityToColor[community] for community in communities
]
for edgeIndex in range(graph.ecount()):
sourceIndex = graph.es[edgeIndex].source
graph.es[edgeIndex]['color'] = graph.vs["color"][sourceIndex] + "20"
fig, ax = plt.subplots(figsize=(10, 10))
drawGraph(graph, ax)
plt.axis("off")
plt.savefig(output_file)
plt.close()
source_by_year = 'data/1991-2019/by_year/dep_*_obstr_0.8_leidenalg'
source_by_mandate = 'data/1991-2019/mandate/dep_*_0.8'
# Called only once
source = 'data/1991-2019/by_year/dep_*_obstr_0.8_leidenalg'
filenames = glob.glob(source + '.xnet')
calculate_dist(filenames)
filenames_by_year = sorted(glob.glob(source_by_year + '_dist.xnet'))
filenames_by_mandate = sorted(glob.glob(source_by_mandate + '_dist.xnet'))
dates_by_year, dates_by_mandate = [], []
nets_by_year, nets_by_mandate = [], []
for filename in filenames_by_year:
net = xnet.xnet2igraph(filename)
net.vs['political_party'] = [
filter_pp_name(p) for p in net.vs['political_party']
]
nets_by_year.append(net.components().giant())
date = int(filename.split('dep_')[1].split('_')[0])
dates_by_year.append(date)
for filename in filenames_by_mandate:
net = xnet.xnet2igraph(filename)
net.vs['political_party'] = [
filter_pp_name(p) for p in net.vs['political_party']
]
nets_by_mandate.append(net.components().giant())