Groupby Pandas会生成多个条件字段

Question

I have a pandas dataframe as such:

df = pandas.DataFrame( { 
    "Label" : ["A", "A", "B", "B", "C" , "C"] , 
    "Value" : [1, 9, 1, 1, 9, 9],
    "Weight" : [2, 4, 6, 8, 10, 12} )

I would like to group the data by 'Label' and generate 2 fields.

• The First field, 'newweight' would sum Weight if Value==1
• The Second field, 'weightvalue' would sum Weight*Value

So I would be left with the following dataframe:

Label     newweight     weightvalue
 A           2               38
 B           14              14
 C           0               198

I have looked into the pandas groupby() function but have had trouble generating the 2 fields with it.

Answer 1

Use groupby.apply , you can do:

df.groupby('Label').apply(
  lambda g: pd.Series({
    "newweight": g.Weight[g.Value == 1].sum(),
    "weightvalue": g.Weight.mul(g.Value).sum()
})).fillna(0)

#       newweight  weightvalue
#Label
#A            2.0         38.0
#B           14.0         14.0
#C            0.0        198.0

Answer 2

pd.DataFrame({'Label':df.Label.unique(),'newweight':df.groupby('Label').apply(lambda x : sum((x.Value==1)*x.Weight)).values,'weightvalue':df.groupby('Label').apply(lambda x : sum(x.Value*x.Weight)).values})
Out[113]: 
  Label  newweight  weightvalue
0     A          2           38
1     B         14           14
2     C          0          198

Answer 3

Fast
Super complicated but very cool approach using Numpy's bincount . And likely very fast.

v = df.Value.values
w = df.Weight.values
p = v * w
f, u = pd.factorize(df.Label.values)

pd.DataFrame(dict(
    newweight=np.bincount(f, p).astype(int),
    weightvalue=np.bincount(f, p * (v == 1)).astype(int)
), pd.Index(u, name='Label'))

       newweight  weightvalue
Label                        
A             38            2
B             14           14
C            198            0

---

Creative
Using pd.DataFrame.eval

e = """
newweight = Value * Weight
weightvalue = newweight * (Value == 1)
"""
df.set_index('Label').eval(e).iloc[:, -2:].sum(level=0)

       newweight  weightvalue
Label                        
A             38            2
B             14           14
C            198            0