CODEWITHSUNDEEP
pythonpandasnumpydataframe
Code_ninja
Code_ninja

Reputation: 117

Adjacency matrix from Pandas edgelist dataframe for undirected graphs

data1 = { 'node1': [1,1,1,2],
     'node2': [2,3,5,4],
     'weight': [1,1,1,1], }
df1 = pd.DataFrame(data1, columns = ['node1','node2','weight'])

I want to create an adjacency matrix from pandas dataframe.The dataframe has edgelist of the undirected graph

OUTPUT:

0 1 1 0 1
1 0 0 1 0
1 0 0 0 0
0 1 0 0 0
1 0 0 0 0

My code:

def adjmat():
    print 'begun creating adjen mat'
    data = sc.loadtxt('training.csv', dtype='str', delimiter=',',skiprows=1)
    data = sc.transpose(data)
    row1 = data[1].astype(int)
    row2 = data[2].astype(int)
    weight=data[3].astype(int)
    n=0
    n1=0
    n2=0


    n1=max(row1)
    n2=max(row2)

    if n1>n2:

        Amat=sc.zeros((n1,n1))
        #matrix=sc.zeros((n1,n1))
        n=n1


    else:
        Amat=sc.zeros((n2,n2))
        #matrix=sc.zeros((n2,n2))
        n=n2


    for i in range(0,len(row1)):

        row=row1[i]
        col=row2[i]

        Amat[row-1][col-1]=weight[i]

    i_lower = np.tril_indices(n, -1)
    Amat[i_lower] = Amat.T[i_lower]



    return Amat

I am looking for code which will be scalable.right now I am deaing with dataset which has 100,000 nodes and this code is not able to handle such large dataset.

Upvotes: 2

Views: 2820

Answers (1)

ubuntu_noob
ubuntu_noob

Reputation: 2365

Using networkx.....

data1 = { 'node1': [1,1,1,2],
 'node2': [2,3,5,4],
 'weight': [1,1,1,1], }
df1 = pd.DataFrame(data1, columns = ['node1','node2','weight'])       
G=nx.from_pandas_dataframe(df1,'node1','node2','weight')

Adjtraining = nx.adjacency_matrix(G)

print Adjtraining.todense()

output

[[0 1 1 0 1]
 [1 0 0 1 0]
 [1 0 0 0 0]
 [0 1 0 0 0]
 [1 0 0 0 0]]

Upvotes: 1

Related Questions