0

I load text file of weighted graph. The text file contains three columns named "FromNodeId", "ToNodeId" and "Sign". Sign is weight of edge. value of Sign is -1 or 1. I want to find the number of output edges (output degree) with the "Sign=1" of each node. Please suggest me a way to solve this problem.

import networkx as nx
G= nx.read_edgelist("soc-sign-epinions.txt",data = [('Sign', int)], create_using=nx.DiGraph())

nodes = G.nodes()
edges = G.edges()
mina
  • 153
  • 6
  • 18
  • 1
    Doesn't `G.degree(node)` give the outdegree if it's a directed graph? – Joel Nov 01 '18 at 02:22
  • I can use ***G.out_degree(node)*** in directed graph. but I want to count out of edges with Sign =1 or Sign =-1 for each node. – mina Nov 01 '18 at 07:43
  • 2
    this would be a bit quicker to answer if it had a [mcve] - in particular, I don't have soc-sign-epinions.txt, so I can't just copy and paste your code. – Joel Nov 01 '18 at 22:25

2 Answers2

1

You can do this directly using pandas. You can read in your data using pd.read_csv('path_to_file'), then filter for edges with positive sign, then groupby origin node and sum up the remaining signs. Here is an example with fake data:

import pandas as pd
data=pd.DataFrame([['a','b',1],
              ['a','c',-1],
              ['a','d',1],
              ['b','a',1],
              ['b','d',-1],
              ['c','a',1],
              ['d','b',1]],
               columns = ["FromNodeId", "ToNodeId","Sign"])
data[data['Sign']==1].groupby('FromNodeId')['Sign'].sum()

returns:

FromNodeId
a    2
b    1
c    1
d    1
Name: Sign, dtype: int64
Johannes Wachs
  • 1,270
  • 11
  • 15
  • 1
    This requires going from the networkx graph to pandas first, which is probably inefficient if the user just wants a single node's information. – Joel Nov 01 '18 at 22:24
  • 1
    If the user just needs this information and is reading in from a txt file, networkx can be avoided entirely. If he or she needs the graph structure then you are right. – Johannes Wachs Nov 02 '18 at 10:41
1

So we'll break this down into two observations. First we can access all the edges out of a node (and the associated data) with G.edges(node, data = True). Second, there are ways to efficiently loop through these edges just counting those with positive sign. More generally, this approach can be used to count the number of edges out of a node that have any particular property.

import networkx as nx


G = nx.DiGraph()
G.add_edge(0,2,sign=-1)
G.add_edge(0,1, sign = 1)
G.add_edge(2,3,sign = 1)
G.add_edge(3,0, sign=-1)
print(G.edges(0, data=True))
>[(0, 2, {'sign': -1}), (0, 1, {'sign': 1})]

Note that the edge (3,0) did not appear here. So G.edges(0, data=True) results in the edges that start from 0 and includes the data you've attached to the edge. (in your final code, obviously you don't actually want this print statement).

Now we'll use that in a generator and sum up the number of elements.

s = sum(1 for (u,v,d) in G.edges(0, data=True) if d['sign']==1)
print(s)
> 1

What I've done is create a generator that goes through all edges out of 0 and if the sign is 1, it adds 1 to the output.

If that last sentence doesn't make sense, look at this answer: https://stackoverflow.com/a/7223557/2966723 to given an idea of what is going on, and for more about generators, start with Understanding generators in Python.

Joel
  • 22,598
  • 6
  • 69
  • 93