Best way(run-time) to aggregate (calculate ratio of) sum to total count based on group by

问题

I'm trying to identify ratio of approved applications(identified by flag '1' and if not then '0') to total applications for each person(Cust_ID). I have achieved this logic by the following code but it takes about 10 mins to compute this for 1.6 M records. Is there a faster to perform the same operation?

# Finding ratio of approved out of total applications
df_approved_ratio = df.groupby('Cust_ID').apply(lambda x:x['STATUS_Approved'].sum()/len(x))

回答1:

I think need aggregate by mean:

df = pd.DataFrame({'STATUS_Approved':[0,1,0,0,1,1],
                   'Cust_ID':list('aaabbb')})

print (df)
   STATUS_Approved Cust_ID
0                0       a
1                1       a
2                0       a
3                0       b
4                1       b
5                1       b

df_approved_ratio = df.groupby('Cust_ID')['STATUS_Approved'].mean()
print (df_approved_ratio)
Cust_ID
a    0.333333
b    0.666667
Name: STATUS_Approved, dtype: float64

print (df.groupby('Cust_ID').apply(lambda x:x['STATUS_Approved'].sum()/len(x)))
Cust_ID
a    0.333333
b    0.666667
Name: STATUS_Approved, dtype: float64

来源：https://stackoverflow.com/questions/51519379/best-wayrun-time-to-aggregate-calculate-ratio-of-sum-to-total-count-based-on