如何在需要两个向量的函数中使用应用程序

Question

I want to pass vector columns to apply in a pandas dataframe, here is a simplification:

def mae(y_true, y_pred):
    return (y_true - y_pred).abs().mean()

df = pd.DataFrame({"y_true": [1.1, 2, 3], "y_pred": [2, 2.5, 3]})
df[df.y_true > 1.5].apply(lambda x: mae(x.y_true, x.y_pred), axis=1)

It gives an error with or without the axis=1 . I want to avoid doing the long way

df_filtered = df[df.y_true > 1.5]
mae(df_filtered.y_true.values, df_filtered.y_pred.values)

It might be something easy, but I've done some research and don't know how to do it. The value returned should be 0.25

Answer 1

You want to avoid using apply for these simple calculations since it is just syntactic sugar for a python level for . Just index your dataframe where df.y_true > 1.5 , and calculate the mae as:

df_ = df[df.y_true > 1.5]
(df_.y_true - df_.y_pred).abs().mean()
# 0.25

Answer 2

I found the piece of code I wanted. Using pipe solves the problem of reassigning:

df[df.y_true > 1.5].pipe(lambda x: mae(x.y_true, x.y_pred))
# 0.25