I'm dipping my toes into network visualizations in Python. I have a dataframe like the following:

| user | nodes    |
| -----| ---------|
| A    | [0, 1, 3]|
| B    | [1, 2, 4]|
| C    | [0, 3]   |
|...   |          |

Is there a way to easily plot a network graph (NetworkX?) from data that contains the list of nodes on each row? The presence of a node in a row would increase the prominence of that node on the graph (or the prominence/weight of the edge in the relationship between two nodes).

I assume some transformation would be required to get the data into the appropriate format for NetworkX (or similar) to be able to create the graph relationships.

Thanks!

I'm dipping my toes into network visualizations in Python. I have a dataframe like the following:

| user | nodes    |
| -----| ---------|
| A    | [0, 1, 3]|
| B    | [1, 2, 4]|
| C    | [0, 3]   |
|...   |          |

Is there a way to easily plot a network graph (NetworkX?) from data that contains the list of nodes on each row? The presence of a node in a row would increase the prominence of that node on the graph (or the prominence/weight of the edge in the relationship between two nodes).

I assume some transformation would be required to get the data into the appropriate format for NetworkX (or similar) to be able to create the graph relationships.

Thanks!

• python
• python-3.x
• pandas
• graph
• networkx

• Your example is unclear. How do you define the edges? For instance why is there no edge between 0 and 1? – mozway Commented 17 hours ago
• That was a mistake! Yes there should be a line 0 to 1 – alpacafondue Commented 15 hours ago
• Got is, thanks for the clarification – mozway Commented 15 hours ago

1 Answer 1

Since you have lists, using pandas would not be more efficient.

You could use itertools to enumerate the edges, and collections.Counter to count them, then build the graph and plot with a width based on the weight:

from itertools import combinations, chain
from collections import Counter
import networkx as nx

c = Counter(chain.from_iterable(combinations(sorted(l), 2) for l in df['nodes']))

G = nx.Graph()
G.add_weighted_edges_from((*e, w) for e, w in c.items())

pos = nx.spring_layout(G)
nx.draw_networkx(G, pos)

for *e, w in G.edges(data='weight'):
    nx.draw_networkx_edges(G, pos, edgelist=[e], width=w)

Output:

Used input:

df = pd.DataFrame({'user': ['A', 'B', 'C'],
                   'nodes': [[0, 1, 3], [1, 2, 4], [0, 3]],
                  })