CODEWITHSUNDEEP
pythonpandasnetworkxgraph-theoryadjacency-matrix
Curious
Curious

Reputation: 815

Networkx does not return a nice graph from adjacency matrix

I have a matrix that is as follows:

adjacency_matrix = [['A', 1, 1, 0, 2], ['B', 1, 1, 1, 3], ['C', 0, 0, 1, 1]]

It shows that A is in "Element 1", "Element 2" but not "Element 3" as it has 1, 1 and 0.

B is "Element 1", "Element 2" and "Element 3" as all values are 1s and etc. The last value is the sum of the 0s and 1s in that sublist.

I created a pandas dataframe to save this into a csv file. Before it saves it, it sorts it by the sum and then drops the last column (sum). 

df = pd.DataFrame(adjacency_matrix, columns = ["Name", "Element 1", "Element 2", "Element 3", "Sum"])
df = df.sort_values(by=['Sum'], ascending=False)
df = df.iloc[:, :-1]

My next step is to use the adjacency matrix and create a nice graph of connections. 

G=from_pandas_edgelist(df, source="Name", target=["Name", "Element 1", "Element 2", "Element 3"])
nx.draw_circular(G, with_labels=True)
plt.axis('equal')
plt.show()

What am I doing wrong? I do not get the undirected graph with "A" connected to both Element 1 and Element 2. I have a feeling my source and target are wrong.

enter image description here

Upvotes: 2

Views: 632

Answers (2)

Chris Adams
Chris Adams

Reputation: 18647

Restructure your adjacency matrix into an edgelist. Here's an example using DataFrame.melt and DataFrame.query:

df = pd.DataFrame(adjacency_matrix, columns = ["Name", "Element 1", "Element 2", "Element 3", "Sum"])
df = df.sort_values(by=['Sum'], ascending=False)
df = df.iloc[:, :-1]

df_edges = (df.melt(id_vars='Name', var_name='target')
            .query('value==1'))

[out]

  Name     target  value
0    A  Element 1      1
1    B  Element 1      1
3    A  Element 2      1
4    B  Element 2      1
7    B  Element 3      1
8    C  Element 3      1


G = nx.from_pandas_edgelist(df_edges, source='Name', target='target')
nx.draw_networkx(G)

enter image description here

Upvotes: 3

CDJB
CDJB

Reputation: 14506

My approach is to restructure your adjacency_matrix to include all pairs:

adjacency_matrix = [['A', 1, 1, 0, 2], ['B', 1, 1, 1, 3], ['C', 0, 0, 1, 1]]

df = pd.DataFrame(adjacency_matrix, columns = ["Name", "Element 1", "Element 2", "Element 3", "Sum"])
df = df.sort_values(by=['Sum'], ascending=False)
df = df.iloc[:, :-1]

df = df.set_index('Name')

edges = df.columns

for i in df.index:
    df[i] = [0 for _ in range(len(df.index))]

for e in edges:
    r = [0 for _ in range(len(df.columns))]
    df.loc[len(df)] = r


as_list = df.index.tolist()
as_list[len(adjacency_matrix):] = edges
df.index = as_list


G=nx.from_pandas_adjacency(df)
nx.draw_circular(G, with_labels=True)
plt.axis('equal')
plt.show()

To make your df as follows:

           Element 1  Element 2  Element 3  B  A  C
B                  1          1          1  0  0  0
A                  1          1          0  0  0  0
C                  0          0          1  0  0  0
Element 1          0          0          0  0  0  0
Element 2          0          0          0  0  0  0
Element 3          0          0          0  0  0  0

Which gives:

enter image description here

Upvotes: 2

Related Questions