def pagerank(self): # from networkx.algorithms.link_analysis import pagerank_scipy # from networkx.algorithms.link_analysis import pagerank_numpy from networkx.algorithms.link_analysis import pagerank from time import time try: start = time() # pagerank(graph, max_iter=1000) # 1.7s for #nodes = 2500 pagerank(self.graph, max_iter=1000, nstart=self.page_ranked) # 0.2-1.5s for #node = 2500 # pagerank_scipy(graph) # 1.0s for #nodes = 2500 # pagerank_numpy(graph) # > 30s if #nodes > 1000 print "Pagerank took: %f seconds" % (time()-start) except ZeroDivisionError: print "ZeroDivisionError in pagerank" page_ranked_sorted = sorted(self.page_ranked.items(), key=lambda x: x[1], reverse=True) print page_ranked_sorted[:4] # from networkx.algorithms.centrality import * # start = time() # degree_centrality = degree_centrality(graph) # 0.003s for 1500 nodes # print "Degree centrality took: %f seconds" % (time()-start) # # start = time() # closeness_centrality = closeness_centrality(graph) # 4s for 1500 nodes # print "Closeness centrality took: %f seconds" % (time()-start) # # start = time() # betweenness_centrality = betweenness_centrality(graph) # 18s for 1500 nodes # print "Betweenness centrality took: %f seconds" % (time()-start) return self.page_ranked
def graph_2():
G = nx.nx.DiGraph()
# G.add_nodes_from([2, 3, 5, 6, 7])
G.add_edges_from([[2, 3], [5, 3], [6, 7], [7, 2], [5, 7]])
# G.add_path([2,3,6,7])
# G.add_path([2,4,5])
# print(list(G.nodes()))
print(list(G.edges()))
print(list(G.out_degree()))
print(list(G.in_degree()))
print(centrality.in_degree_centrality(G))
print(link_analysis.pagerank(G, personalization={2: -4}))
print(link_analysis.pagerank(G, dangling={5: 0, 7: 1}))
def rank_it(self):
rank = la.pagerank(self.pg.graph, max_iter=self.simulations, alpha=self.p_restart)
print "\t".join(["id","pagerank","freq-abs","freq-rel","centrality"])
for k, v in sorted(rank.items(), key=lambda q: q[1], reverse=True):
print "%s:\t%s\t%s\t%s\t%s" % (k, v,
self.pg.freq[k],
self.pg.freq[k]/float(self.pg.total),
self.pg.central[k])
def pagerank(self):
# from networkx.algorithms.link_analysis import pagerank_scipy
# from networkx.algorithms.link_analysis import pagerank_numpy
from networkx.algorithms.link_analysis import pagerank
from time import time
try:
start = time()
# pagerank(graph, max_iter=1000) # 1.7s for #nodes = 2500
pagerank(self.graph, max_iter=1000,
nstart=self.page_ranked) # 0.2-1.5s for #node = 2500
# pagerank_scipy(graph) # 1.0s for #nodes = 2500
# pagerank_numpy(graph) # > 30s if #nodes > 1000
print "Pagerank took: %f seconds" % (time() - start)
except ZeroDivisionError:
print "ZeroDivisionError in pagerank"
page_ranked_sorted = sorted(self.page_ranked.items(),
key=lambda x: x[1],
reverse=True)
print page_ranked_sorted[:4]
# from networkx.algorithms.centrality import *
# start = time()
# degree_centrality = degree_centrality(graph) # 0.003s for 1500 nodes
# print "Degree centrality took: %f seconds" % (time()-start)
#
# start = time()
# closeness_centrality = closeness_centrality(graph) # 4s for 1500 nodes
# print "Closeness centrality took: %f seconds" % (time()-start)
#
# start = time()
# betweenness_centrality = betweenness_centrality(graph) # 18s for 1500 nodes
# print "Betweenness centrality took: %f seconds" % (time()-start)
return self.page_ranked
def compute_pagerank(self):
self.kg = KG(self.train_facts,
entity_num=len(self.entity_dict),
relation_num=len(self.relation_dict))
graph = networkx.DiGraph(self.kg.to_networkx())
print("Begin to compute pagerank")
self.pagerank = pagerank(graph)
self.pagerank = [
self.pagerank[entity] for entity in range(len(self.pagerank))
]
print("Begin to save pagerank")
with open(os.path.join(self.data_directory, "pagerank.txt"),
"w") as output:
for value in self.pagerank:
output.write("{}\n".format(value))
print("Complete save pagerank")
def rank_it(self, out):
f_out = open(out, "w")
rank = la.pagerank(self.graph, personalization=self.personalize,
max_iter=self.simulations, alpha=self.p_restart)
size_of_rank = len(rank)
f_out.write("\t".join(["id", "pagerank", "odds", "freq-abs", "freq-rel"]) + "\n")
for k, v in sorted(rank.items(), key=lambda q: q[1],
reverse=True):
f_out.write("%s:\t%s\t%s\t%s\t%s\n" % (k, v,
float(v) * float(size_of_rank),
self.freq[k],
self.freq[k]/float(self.total)))
f_out.close()
def pagerank_worker(graph, page_ranked):
print "Pagerank on graph with %d nodes and %d edges." \
% (len(graph.nodes()), \
len(graph.edges()))
for node in graph.nodes():
page_ranked.setdefault(node, 1)
from networkx.algorithms.link_analysis import pagerank
from time import time
try:
start = time()
page_ranked = pagerank(graph, max_iter=1000, nstart=page_ranked) # 0.2-1.5s for #node = 2500
print "Pagerank took: %f seconds" % (time()-start)
except ZeroDivisionError:
print "ZeroDivisionError in pagerank"
page_ranked_sorted = sorted(page_ranked.items(), key=lambda x: x[1], reverse=True)
print page_ranked_sorted[:4]
def pagerank_worker(graph, page_ranked):
print "Pagerank on graph with %d nodes and %d edges." \
% (len(graph.nodes()), \
len(graph.edges()))
for node in graph.nodes():
page_ranked.setdefault(node, 1)
from networkx.algorithms.link_analysis import pagerank
from time import time
try:
start = time()
page_ranked = pagerank(graph, max_iter=1000,
nstart=page_ranked) # 0.2-1.5s for #node = 2500
print "Pagerank took: %f seconds" % (time() - start)
except ZeroDivisionError:
print "ZeroDivisionError in pagerank"
page_ranked_sorted = sorted(page_ranked.items(),
key=lambda x: x[1],
reverse=True)
print page_ranked_sorted[:4]
with open(graph_dump_filename, "r") as dump:
graph = cPickle.load(dump)
return graph
if __name__ == "__main__":
G = get_graph()
result = {"pagerank_ids": [], "harmonic_ids": []}
logger.info('graph nodes %d' % G.number_of_nodes())
logger.info('harmonic')
harmonic = hc(G)
sorted_h = sorted(harmonic.items(),
key=operator.itemgetter(1),
reverse=True)
result['harmonic_ids'] = map(lambda x: str(x[0]), sorted_h[0:200])
logger.info('pr')
pr = pagerank(G, alpha=0.8507246376811566)
sorted_pr = sorted(pr.items(), key=operator.itemgetter(1), reverse=True)
result['pagerank_ids'] = map(lambda x: str(x[0]), sorted_pr[0:200])
with open('lab6centralities.json', 'w') as f:
f.write(json.dumps(result))
f.close()
logger.info('finish')
