1

Is there a way in Pandas to number groups in a DataFrame, based on column values? If my frame looks like this

  Column1 Column2  Column3
0       A       X       23
1       A       X       45
2       A       Y       32
3       A       Y       53
4       A       Y       67
5       B       X       85
6       B       Y       12
7       B       Y       94

What I'd like to be able to do is something like 

df.group_numbers(['Column1', 'Column2'])

  Column1 Column2  Column3  GroupNumber
0       A       X       23            1
1       A       X       45            1
2       A       Y       32            2
3       A       Y       53            2
4       A       Y       67            2
5       B       X       85            3    
6       B       Y       12            4
7       B       Y       94            4
Gree Tree Python
  • 529
  • 1
  • 6
  • 22
  • 1
    This is a bit like a multi-column factorize: http://stackoverflow.com/questions/16453465/multi-column-factorize-in-pandas – Alex Riley Oct 30 '15 at 19:59

1 Answers1

1

As suggested in ajcr's comment, pd.factorize is the way to go. In your case you can add the two columns to quickly create an array of keys by adding the two columns with some delimiter between. The delimiter is to avoid confusing pairs such as ab, c and a, bc as suggested by DSM. 

df['GroupNumber'] = pd.factorize(df.Column1 + ' ' + df.Column2)

It's still faster than using pd.lib.fast_zip.

JoeCondron
  • 8,546
  • 3
  • 27
  • 28