如何为熊猫 groupby 使用多个 lambda 函数

Question

I'm calculating various evaluation metric for a dataframe with several group. Here's my code

import pandas as pd
from sklearn.metrics import mean_squared_error, median_absolute_error
temp = pd.DataFrame({"group":['A','A','B','B'],"actual":[10,2,3,4],"pred":[0,1,2,3]})

temp.groupby("group").apply(lambda x : mean_squared_error(x['actual'],x['pred'])).to_frame('MSE').reset_index()

    group   MSE
0   A       50.5
1   B       1.0

temp.groupby("group").apply(lambda x : median_absolute_error(x['actual'],x['pred'])).to_frame('MAE').reset_index()

group   MAE
0   A   5.5
1   B   1.0

If I have 5 metrics, than I need to write groupby + apply 5 times, or execute is as a loop. But is there any native way from pandas to call for multiple apply in a single groupby object ?

Maybe something like this :

temp.groupby("group").agg({"MAE": lambda x : median_absolute_error(x['actual'],x['pred']),"MSE": lambda x : mean_squared_error(x['actual'],x['pred'])})

    group  MSE  MAE
0   A      50.5 5.5
1   B      1.0  1.0

the code is wrong but I think you get what i'm trying to do.

Answer 1

Try returning a Series from groupby apply :

new_df = temp.groupby("group", as_index=False).apply(
    lambda x: pd.Series({'MSE': mean_squared_error(x['actual'], x['pred']),
                         'MAE': median_absolute_error(x['actual'], x['pred'])})
)

new_df :

  group   MSE  MAE
0     A  50.5  5.5
1     B   1.0  1.0