How do you find the top 3 probabilities from 10 columns in a row in Pandas DataFrame?

Question

I have a Pandas Dataframe with a probability for each column and I have 10 columns. Row represents a record. I'd like to select top 3 probabilities and return that column names (kind of like argmax).

Since I have a large volume of rows, I don't want to use loop hopefully. Could someone help me do this?

The DataFrame looks like this:

    col0    col1    col2    col3    col4    col5    col6    col7    col8    col9    
0   0.089659    0.0 0.0 0.0 0.228481    0.000000    0.0 0.575523    0.0 0.089667    
1   0.000000    0.0 0.0 0.0 0.000000    0.422125    0.0 0.516324    0.0 0.000000    
2   0.000000    0.0 0.0 0.0 0.000000    0.304416    0.0 0.659216    0.0 0.000000    
3   0.598459    0.0 0.0 0.0 0.100792    0.082668    0.0 0.144577    0.0 0.053502    
4   0.100000    0.1 0.1 0.1 0.100000    0.100000    0.1 0.100000    0.1 0.100000

Could you provide an example of your DataFrame? – John Karasinski Jun 05 '16 at 03:22 — John Karasinski, Jun 05 '16 at 03:22
thanks for asking, karasinsk! I added my example dataframe – user3368526 Jun 05 '16 at 03:38 — user3368526, Jun 05 '16 at 03:38
could you also post a desired output/result set? – MaxU - stand with Ukraine Jun 05 '16 at 07:50 — MaxU - stand with Ukraine, Jun 05 '16 at 07:50

score 2 · Answer 1 · answered Jun 05 '16 at 03:51

Given the df

df
Out[145]: 
       col0  col1  col2  col3      col4      col5  col6      col7  col8  \
0  0.089659   0.0   0.0   0.0  0.228481  0.000000   0.0  0.575523   0.0   
1  0.000000   0.0   0.0   0.0  0.000000  0.422125   0.0  0.516324   0.0   
2  0.000000   0.0   0.0   0.0  0.000000  0.304416   0.0  0.659216   0.0   
3  0.598459   0.0   0.0   0.0  0.100792  0.082668   0.0  0.144577   0.0   
4  0.100000   0.1   0.1   0.1  0.100000  0.100000   0.1  0.100000   0.1   

       col9  
0  0.089667  
1  0.000000  
2  0.000000  
3  0.053502  
4  0.100000

You can use nlargest:

import numpy as np

x=pd.DataFrame(df).T

rslt = pd.DataFrame(np.zeros((0,3)), columns=['top1','top2','top3'])
for i in x.columns:
    df1row = pd.DataFrame(x.nlargest(3, i).index.tolist(), index=['top1','top2','top3']).T
    rslt = pd.concat([rslt, df1row], axis=0)

print rslt

   top1  top2  top3
0  col7  col4  col9
0  col7  col5  col0
0  col7  col5  col0
0  col0  col7  col4
0  col0  col1  col2

@ToussaintLouverture, Thank you for liking the answer! I'm flattered! — MaThMaX, Feb 06 '18 at 06:30

score 0 · Answer 2 · answered Jun 05 '16 at 03:45

0

You should look into nlargest. This might be enough to get you going:

df.T.nlargest(3, df.T.columns).T

answered Jun 05 '16 at 03:45

John Karasinski

977
7
16

@downvoters, could you please explain why has it been downvoted? – MaxU - stand with Ukraine Jun 05 '16 at 07:51

How do you find the top 3 probabilities from 10 columns in a row in Pandas DataFrame?

2 Answers2