def do_link_prediction(
NETWORK_FILE,
TEST_EDGE_FILE,
NEGATIVE_EDGE_FILE,
DIRECTED,
OUT_FILE,
DIM,
NUMBER_OF_CORES,
):
G = read_graph(NETWORK_FILE, directed=DIRECTED)
model = node2vec.Node2Vec(G,
directed=DIRECTED,
dimensions=DIM,
workers=NUMBER_OF_CORES)
model.simulate_walks()
emb = model.learn_embedding()
test_edges = read_edge_file(TEST_EDGE_FILE)
negative_edges = read_edge_file(NEGATIVE_EDGE_FILE)
name = "\t".join([NETWORK_FILE, str(DIM)])
test = LinkPredictionTask(test_edges,
negative_edges,
emb,
name=name,
is_persona_emb=False)
test.do_link_prediction()
test.write_result(OUT_FILE)
def do_link_prediction(
NETWORK_FILE,
TEST_EDGE_FILE,
NEGATIVE_EDGE_FILE,
DIRECTED,
OUT_FILE,
LAMBDA,
DIM,
NUMBER_OF_CORES,
):
G = read_graph(NETWORK_FILE, directed=DIRECTED)
model = Persona2Vec(G,
lambd=LAMBDA,
directed=DIRECTED,
dimensions=DIM,
workers=NUMBER_OF_CORES)
emb = model.embedding
test_edges = read_edge_file(TEST_EDGE_FILE)
negative_edges = read_edge_file(NEGATIVE_EDGE_FILE)
name = "\t".join([NETWORK_FILE, str(LAMBDA), str(DIM)])
test = LinkPredictionTask(
test_edges,
negative_edges,
emb,
name=name,
is_persona_emb=True,
node_to_persona=model.node_to_persona,
)
test.do_link_prediction()
test.write_result(OUT_FILE)
def main():
"""
Parsing command line parameters.
Reading data, embedding base graph, creating persona graph and learning a splitter.
saving the persona mapping and the embedding.
"""
args = parse_args()
tab_printer(args)
G = read_graph(args.input, args.weighted, args.directed)
model = Persona2Vec(
G,
lambd=args.lambd,
clustering_method="connected_component",
directed=args.directed,
num_walks_base=args.num_walks_base,
walk_length_base=args.walk_length_base,
window_size_base=args.window_size_base,
num_walks_persona=args.num_walks_persona,
walk_length_persona=args.walk_length_persona,
window_size_persona=args.window_size_persona,
p=args.p,
q=args.q,
dimensions=args.dimensions,
epoch_base=args.epoch_base,
epoch_persona=args.epoch_persona,
workers=args.workers,
)
model.save_persona_network(args.persona_network)
model.save_persona_to_node_mapping(args.persona_to_node)
model.save_node_to_persona_mapping(args.node_to_persona)
model.save_base_embedding(args.base_emb)
model.save_persona_embedding(args.persona_emb)
def train_test_set_split(IN_FILE, INDEX, DIRECTED):
G = read_graph(IN_FILE, directed=DIRECTED)
OUTPUT_PATH = "{}_{}".format(IN_FILE.split(".")[0], INDEX)
splitter = NetworkTrainTestSplitter(G, directed=DIRECTED)
splitter.train_test_split()
splitter.generate_negative_edges()
splitter.save_splitted_result(OUTPUT_PATH)
def do_link_prediction(
NETWORK_FILE,
TEST_EDGE_FILE,
NEGATIVE_EDGE_FILE,
DIRECTED,
OUT_FILE,
DIM,
NUMBER_OF_CORES,
):
G = read_graph(NETWORK_FILE, directed=DIRECTED)
model = node2vec.Node2Vec(G,
directed=DIRECTED,
dimensions=DIM,
workers=NUMBER_OF_CORES)
model.simulate_walks()
emb = model.learn_embedding()
test_edges = read_edge_file(TEST_EDGE_FILE)
negative_edges = read_edge_file(NEGATIVE_EDGE_FILE)
test_hardmord_vectors = [
np.multiply(emb[src], emb[tag]) for src, tag in test_edges
]
negative_hardmord_vectors = [
np.multiply(emb[src], emb[tag]) for src, tag in negative_edges
]
xs = np.concatenate([test_hardmord_vectors, negative_hardmord_vectors])
ys = np.concatenate([
np.ones(len(test_hardmord_vectors)),
np.zeros(len(negative_hardmord_vectors))
])
clf = LogisticRegression(random_state=0).fit(xs, ys)
predicted_ys = clf.predict(xs)
ROC_AUC_value = roc_auc_score(ys, predicted_ys)
name = "\t".join([NETWORK_FILE, str(DIM)])
f = open(OUT_FILE, "a")
f.write("{}\t{}\n".format(*[name, str(ROC_AUC_value)]))
f.close()
def convert_network_files_for_splitter(NETWORK_FILE, CONVERTED_NETWORK_FILE, DIRECTED):
G = read_graph(NETWORK_FILE, directed=DIRECTED)
G_ = nx.convert_node_labels_to_integers(G, label_attribute="old_label")
nx.write_edgelist(G_, CONVERTED_NETWORK_FILE)
return G_