从 one-hot 编码的 dataframe 中获得胜率

Question

i have this one-hot encoded dataframe the win column is one if the observations won and 0 otherwise. how do i get the win percentage for each one of the categories (food,fitness,retail,grocery)? each observation can be in multiple categories and some ids are duplicated because each id is an experiment where multiple things can be tested.

id food    fitness   retail    grocery win
1  1       0         1         1       1
2  1       0         0         0       0
3  0       1         0         0       1
4  1       0         0         1       1
4  1       0         0         1       0
5  1       0         1         0       1
6  0       1         1         0       1
6  0       1         1         0       0

expected output

category win_percentage
food     .6
fitness  .66
retail   .75
grocery  .66

Answer 1

You can do this:

df1 = df.drop(['id','win'], 1)
win_percent = df1[df1 == 1].mul(df.win, 0).mean()

Output

win_percent

food       0.600000
fitness    0.666667
retail     0.750000
grocery    0.666667
dtype: float64

---

To get exactly your expected output:

win_percent.to_frame('win_percent').rename_axis('category').reset_index()

  category  win_percent
0     food     0.600000
1  fitness     0.666667
2   retail     0.750000
3  grocery     0.666667

Answer 2

category, win_perecentage = [], []
for c in df.loc[:, "food":"grocery"]:
    category.append(c)
    win_perecentage.append(df.loc[df[c] == 1, "win"].sum() / df[c].sum())

df_out = pd.DataFrame(
    {"category": category, "win_perecentage": win_perecentage}
)
print(df_out)

Prints:

  category  win_perecentage
0     food         0.600000
1  fitness         0.666667
2   retail         0.750000
3  grocery         0.666667