Python：如何从csv文件创建图形节点和边？

Question

I have a csv data file where each row denotes one event. A simplified example would be,

  Datetime                 ColA     ColB    ColC   
  2015/07/12 08:45:34      ABC       12      
  2015/07/12 08:46:04      DCD       10     ABC 
  2015/07/12 08:46:23      XYZ       34     ABC 
  2015/07/12 08:46:56      MNO       10     XYZ
  2015/07/12 08:46:56      FGH       20     

So, each row will be a node with properties denoted by values of each columns Datetime, ColA, ColB, ColC . Each of these nodes is connected by a relationship between ColA and ColC .

So, in this example there is an edge from row 1 to row 2 and row 3, since ColC of the latter rows is equal to ColA of the first row. Row 3 and row 4 are similarly connected by an directed edge.

Row 1 has no ColC so it is not connected to any nodes higher up, so is row 5.

How can I create a graph data structure to create this relationship in Python? They should all be ordered chronologically, and in case there are two rows with ColA that matches a rows ColC , the one closer in time is chosen.

Answer 1

You could build a bipartite graph with the datetime as one part and the ColA/ColC values as the other. Then "project" the graph onto the datetime nodes - create a link between two datetimes if they both link to a ColA/ColC node.

Here is some code that shows one way to do that to create an undirected graph. I didn't understand what the directions meant in your example.

import csv
import StringIO
import networkx as nx
from networkx.algorithms import bipartite

data ="""Datetime,ColA,ColB,ColC
2015/07/12 08:45:34,ABC,12,
2015/07/12 08:46:04,DCD,10,ABC
2015/07/12 08:46:23,XYZ,34,ABC
2015/07/12 08:46:56,MNO,10,XYZ
2015/07/12 08:46:56,FGH,20,"""

G = nx.Graph()
csvfile = StringIO.StringIO(data)
reader = csv.DictReader(csvfile)
nodes = []
for row in reader:
    nodes.append(row['Datetime'])
    G.add_node(row['Datetime'])
    if row['ColA'] != '':
        G.add_edge(row['Datetime'],row['ColA'])
    if row['ColC'] != '':
        G.add_edge(row['Datetime'],row['ColC'])
print G.edges()
B = bipartite.projected_graph(G, nodes)
print B.edges()

OUTPUT

[('2015/07/12 08:46:23', 'XYZ'), ('2015/07/12 08:46:23', 'ABC'), ('ABC', '2015/07/12 08:46:04'), ('ABC', '2015/07/12 08:45:34'), ('DCD', '2015/07/12 08:46:04'), ('FGH', '2015/07/12 08:46:56'), ('2015/07/12 08:46:56', 'XYZ'), ('2015/07/12 08:46:56', 'MNO')]
[('2015/07/12 08:46:23', '2015/07/12 08:46:04'), ('2015/07/12 08:46:23', '2015/07/12 08:46:56'), ('2015/07/12 08:46:23', '2015/07/12 08:45:34'), ('2015/07/12 08:46:04', '2015/07/12 08:45:34')]