Load graph from pandas dataframe as adjacency matrix (from https://www.bountysource.com/issues/31619133-load-graph-from-pandas-dataframe-as-adjacency-matrix)

pandas_matrix.py

import numpy as np
# Assume that we have a dataframe df which is an adjacency matrix

def find_edges(df):
    """Finds the edges in the square adjacency matrix, using
    vectorized operations. Returns a list of pairs of tuples
    that represent the edges."""

    values = df.values  # Adjacency matrix of 0's and 1's
    n_rows, n_columns = values.shape
    indices = np.arange(n_rows*n_columns)
    values = values.flatten()
    _indices = indices[values == 1]  # A value of 1 means that the edge exists

    # Create two arrays `rows` and `columns` such that for an edge i,
    # (rows[i], columns[i]) is its coordinate in the df
    rows = _indices / n_columns  
    columns = _indices % n_columns

    # Convert the coordinates to actual names
    row_names = df.index[rows]
    column_names = df.columns[columns]
    return zip(row_names, column_names)    # Possible that itertools.izip is faster

G = nx.DiGraph()
G.add_nodes_from(df.index.tolist())

edges = find_edges(df)
G.add_edges_from(edges)  # Speed is questionable, not sure if this is vectorized