python中相似矩阵的Networkx图

Question

i am trying the library networkx for the first time trying to plot a graph using a disimilarity matrix i made which has around 500 elements,

array([[0.        , 0.56666667, 0.41666667, ..., 0.8037037 , 0.89814815,
        0.86296296],
       [0.56666667, 0.        , 0.35833333, ..., 0.73703704, 0.91481481,
        0.87037037],
       [0.41666667, 0.35833333, 0.        , ..., 0.8037037 , 0.98148148,
        0.86296296],
       ...,
       [0.8037037 , 0.73703704, 0.8037037 , ..., 0.        , 0.67777778,
        0.85740741],
       [0.89814815, 0.91481481, 0.98148148, ..., 0.67777778, 0.        ,
        0.78518519],
       [0.86296296, 0.87037037, 0.86296296, ..., 0.85740741, 0.78518519,
        0.        ]])

and i want to find a good visual way to show relationships based on distance between them

i tried

import networkx as nx


G = nx.from_numpy_matrix(dismatrix)
G = nx.relabel_nodes(G, dict(zip(range(len(G.nodes())),string.ascii_uppercase)))  
nx.draw(G, with_labels=True, font_weight='bold')

and the result i found isn't that great .. do you guys have any ways how to improve it or actually show something out of it ?

graph ploted

Answer 1

You have several options. First you could apply a community detection algorithm that uses your edge weights (dissimilarity) to find groups of similar nodes. These labels can then be used to colour your nodes.

Second, you could simply trim edges using a threshold - such as keep only those edges where similarity >= 0.75.

In either case you then need to look at what networkx layout you are using in your plots. These include: spring, spectral, force directed , Circular etc. Once you have your graph It might be easier to create a “nice” visual in graph plotting software such as Gephi.