Fill missing values in pandas dataframe

user_10487326 0
Viewed 72

I have a pandas dataframe with two columns : locationid, geo_loc. locationid column has missing values.

I want to get the geo_loc value of the missing locationid row, then search this geo_loc value in geo_loc column and get the loction id.

df1 = pd.DataFrame({'locationid':[111, np.nan, 145, np.nan, 189,np.nan, 158, 145],
                     'geo_loc':['G12','K11','B16','G12','B22','B16', 'K11',he l 'B16']})
df

enter image description here

I need the final output like this:

enter image description here

index 1 of locationid is missing and the corresponding geo_loc value is 'K11'. I would look this 'K11' in geo_loc column and index 6 has locationid 158. With this value I want to fill the missing value in index 1.

I tried these codes and they didnt work.

df1['locationid'] = df1.locationid.fillna(df1.groupby('geo_loc')['locationid'].max())

df1['locationid'] = df1.locationid.fillna(df1.groupby('geo_loc').apply(lambda x: print(list(x.locationid)[0])))
data-science pandas machine-learning dataframe python
1 Answers

Use GroupBy.transform for Series with same size like original filled by aggregate values max:

df1['locationid']=df1.locationid.fillna(df1.groupby('geo_loc')['locationid'].transform('max'))
print (df1)
   locationid geo_loc
0       111.0     G12
1       158.0     K11
2       145.0     B16
3       111.0     G12
4       189.0     B22
5       145.0     B16
6       158.0     K11
7       145.0     B16

If values are strings is is possible by trick - remove missing values with Series.dropna in lambda function, strings are compared lexicographically:

df1 = pd.DataFrame({'locationid':[111, np.nan, 145, np.nan, 189,np.nan, 158, 145],
                     'geo_loc':['G12','K11','B16','G12','B22','B16', 'K11', 'B16']})

#sample data strings with missing values
df1['locationid'] = df1['locationid'].dropna().astype(str) + 'a'


df1['locationid']= (df1.groupby('geo_loc')['locationid']
                       .transform(lambda x: x.fillna(x.dropna().max())))

print (df1)
  locationid geo_loc
0     111.0a     G12
1     158.0a     K11
2     145.0a     B16
3     111.0a     G12
4     189.0a     B22
5     145.0a     B16
6     158.0a     K11
7     145.0a     B16
user_29010020
Related