欧氏距离的python数据框矩阵

Question

I would like to create an own customized k nearest neighbor method.

For this I would need a matrix (x : y) which returns the distance for each combination of x and y for a given function (eg euclidean based on 7 items of my dataset).

eg

data:
   x1  x2  x3
  row 1:  1   2   3
  row 2:  1   1   1 
  row 3:  4   2   3

if I select x1 and x2 and euclidean, then the output should be a 3x3 output

1:1=0
1:2 =sqrt((1-1)^2+(2-1)^2)=1
1:3 =sqrt((1-4)^2+(2-2)^2)=sqrt(3)
2:1=1:2=1
2:2=0
2:3=sqrt((1-4)^2+(1-2)^2)=2
3:3=0

and so forth...

how to write that without iterating through the dataframe?

Thanks in advance for your support.

Answer 1

You can use scipy.spatial.distance.pdist and scipy.spatial.distance.squareform :

from scipy.spatial.distance import pdist, squareform

dist = pdist(df[['x1', 'x2']], 'euclidean')
df_dist = pd.DataFrame(squareform(dist))

If you just want an array as your output, and not a DataFrame, just use squareform by itself, without wrapping it in a DataFrame.

The resulting output (as a DataFrame):

     0         1         2
0  0.0  1.000000  3.000000
1  1.0  0.000000  3.162278
2  3.0  3.162278  0.000000