CODEWITHSUNDEEP
pandasdataframeapplynumber-formattingnumeric
Nik Tur
Nik Tur

Reputation: 53

Converting only specific columns in dataframe to numeric

I currently have a dataframe with n number of number-value columns and three columns that are datetime and string values. I want to convert all the columns (but three) to numeric values but am not sure what the best method is. Below is a sample dataframe (simplified):

df2 = pd.DataFrame(np.array([[1, '5-4-2016', 10], [1,'5-5-2016', 5],[2, '5- 
4-2016', 10], [2, '5-5-2016', 7], [5, '5-4-2016', 8]]), columns= ['ID', 
'Date', 'Number'])

I tried using something like (below) but was unsuccessful. 

exclude = ['Date']
df = df.drop(exclude, 1).apply(pd.to_numeric, 
errors='coerce').combine_first(df)

The expected output: (essentially, the datatype of fields 'ID' and 'Number' change to floats while 'Date' stays the same)

        ID     Date         Number 
 0      1.0    5-4-2016       10.0  
 1      1.0    5-5-2016       5.0   
 2      2.0    5-4-2016       10.0 
 3      2.0    5-5-2016       7.0
 4      5.0    5-4-2016       8.0

Upvotes: 0

Views: 996

Answers (2)

Andy L.
Andy L.

Reputation: 25239

You need to call to_numeric with option downcast='float', if you want it change to float. Otherwise, it will be int. You also need to join back to non-converted columns of the original df2

df2[exclude].join(df2.drop(exclude, 1).apply(pd.to_numeric, downcast='float', errors='coerce'))

Out[1815]:
       Date   ID  Number
0  5-4-2016  1.0    10.0
1  5-5-2016  1.0     5.0
2  5-4-2016  2.0    10.0
3  5-5-2016  2.0     7.0
4  5-4-2016  5.0     8.0

Upvotes: 0

Shrinath Viswanathan
Shrinath Viswanathan

Reputation: 131

Have you tried Series.astype()?

df['ID'] = df['ID'].astype(float)
df['Number'] = df['Number'].astype(float)

or for all columns besides date:

for col in [x for x in df.columns if x != 'Date']:
   df[col] = df[col].astype(float)

or

df[[x for x in df.columns if x != 'Date']].transform(lambda x: x.astype(float), axis=1)

Upvotes: 1

Related Questions