Reference:

node2vec: Scalable Feature Learning for Networks. Aditya Grover and Jure Leskovec. Knowledge Discovery and Data Mining, 2016.

https://github.com/aditya-grover/node2vec

https://github.com/lucashu1/link-prediction

updating node2vec from python 2 to python 3:

1. In node2vecLinkPred.py,

   1.1 change "walks = [map(str, walk) for walk in walks]" to "walks = [list(map(str, walk)) for walk in walks]"

   1.2 change "LogisticRegression(random_state=0)" to "LogisticRegression(random_state=0,solver='lbfgs')"

2. In generateGraphPickle.py,

   2.1 we can read_edgelist directly using "edges_dir" rather than open it to a file:
   g = nx.read_edgelist(edges_dir,nodetype=int)

   2.2 features_series.as_matrix() --> features_series.values

   2.3 df = pd.read_table(feats_dir, sep=' ', header=None, index_col=0) --> df = pd.read_csv(feats_dir, sep=' ', header=None, index_col=0)