I have a pandas DataFrame containing two columns [‘A’, ‘B’]. Each column is made up of integers.
I want to construct a sparse matrix with the following properties:
- row index is all integers from 0 to the max value in the dataframe
- column index is the same as row index
- entry i,j = 1 if [i,j] or [j,i] is a row of my dataframe (1 should be the max value of the matrix).
Most importantly, I want to do this using
coo_matrix((data, (i, j)))
from scipy.sparse as I’m trying to understand this constructor and this particular way of using it. I have never worked with sparse matrices before. I’ve tried a few things but none of them is working.
EDIT
Sample code
Defining the dataframe
In [96]: df = pd.DataFrame(np.random.randint(5, size=(10,2))) In [97]: df.columns = ['a', 'b'] In [98]: df Out[98]: a b 0 0 3 1 1 4 2 3 3 3 2 0 4 0 2 5 1 0 6 1 1 7 2 3 8 3 4 9 3 2
The closest I’ve come to a solution
In [100]: scipy.sparse.coo_matrix((np.ones_like(df['a']), (df['a'].array, df['b' ...: ].array))).toarray() Out[100]: array([[0, 0, 1, 1, 0], [1, 1, 0, 0, 1], [1, 0, 0, 1, 0], [0, 0, 1, 1, 1]])
The problem is this isn’t a symmetric matrix (as it doesn’t add to both i,j and j,i for a given row) and I think it would give values greater than 1 if there were duplicate rows.
Advertisement
Answer
import numpy as np import pandas as pd from scipy.sparse import coo_matrix df = pd.DataFrame(np.random.default_rng(seed=100).integers(5, size=(10,2))) df.columns = ['a', 'b'] arr = coo_matrix((np.ones_like(df.a), (df.a.values, df.b.values)))
This is what you’ve got. It gives you i,j >= 1 if [i,j] is in df.
arr = arr + arr.T array([[0, 1, 2, 2, 0], [1, 0, 0, 0, 0], [2, 0, 0, 1, 2], [2, 0, 1, 0, 1], [0, 0, 2, 1, 2]])
Now i,j >= 1 if [i,j] or [j,i] is in df.
arr.data = np.ones_like(arr.data)
Now i,j = 1 if [i,j] or [j,i] is in df.
array([[0, 1, 1, 1, 0], [1, 0, 0, 0, 0], [1, 0, 0, 1, 1], [1, 0, 1, 0, 1], [0, 0, 1, 1, 1]])