CODEWITHSUNDEEP
pythonpandas
tam63
tam63

Reputation: 313

Pandas groupby on mixed type column

I am trying to apply pandas groupby to a column which contains floats and strings. The DataFrame looks like:

     name      value
0    var_1     1.4
1    var_2     1110
3    var_2     900
4    var_3     'some_str'
5    var_1     2.7

I am trying to apply the groupby method so that the output dataframe looks something like:

     name      value
0    var_1     2.15
1    var_2     1005
2    var_3     'some_str'

i.e. the get the average of all those values which are recorded multiple times, and preserve the no-numeric values as they are.

If the column was only made up of numeric types this would be simple enough to implement as:

new_df = df.groupby('name').mean().reset_index()

Is there an easy way to overcome the mixed types which makes this method as I have written above inapplicable?

Upvotes: 1

Views: 475

Answers (1)

jezrael
jezrael

Reputation: 863226

Use try-except statement:

#if need convert strings column to mixed values
df['value'] = pd.to_numeric(df['value'], errors='coerce').fillna(df['value'])

def f(x):
    try:
        return x.mean()
    except:
        return ','.join(x)

new_df = df.groupby('name')['value'].apply(f).reset_index()
print (new_df)
    name       value
0  var_1        2.05
1  var_2        1005
2  var_3  'some_str'

Upvotes: 2

Related Questions