I have a dataframe and would like to add sums of specific rows into this dataframe. So for example I have
df = pd.DataFrame({'prod':['a','a','a','b','b','b','c','c','c'], 'attribute':['x','y','z','x','y','z','x','y','z'],
'number1':[1,2,2,3,4,3,5,1,1], 'number2':[10,2,3,3,1,2,3,1,1], 'number3':[1,4,3,5,7,1,3,0,1]})
How can I add for each prod a, b and c the sum of number 1/2/3 of the attributes y and z as a new row? So it looks like this
prod attribute number1 number2 number3
0 a x 1 10 1
1 a y 2 2 4
2 a z 2 3 3
3 a sum_yz 4 5 7
4 b x 3 3 5
5 b y 4 1 7
6 b z 3 2 1
7 b sum_yz 7 3 8
8 c x 5 3 3
9 c y 1 1 0
10 c z 1 1 1
11 c sum_yz 2 2 1
You need concat
with a condtional groupby
.
You can filter the dataframe by using isin
and add a new column with assign
.
First let's select the target cols to sum.
cols = [col for col in df.columns if 'number' in col]
df1 = pd.concat(
[
df,
df[df["attribute"].isin(["y", "z"])]
.groupby("prod")[cols]
.sum()
.assign(attribute="sum_yz")
.reset_index(),
]
).sort_values("prod")
print(df1)
prod attribute number1 number2 number3
0 a x 1 10 1
1 a y 2 2 4
2 a z 2 3 3
0 a sum_yz 4 5 7
3 b x 3 3 5
4 b y 4 1 7
5 b z 3 2 1
1 b sum_yz 7 3 8
6 c x 5 3 3
7 c y 1 1 0
8 c z 1 1 1
2 c sum_yz 2 2 1
You could make a separate DataFrane and append it back to the original DataFrame, something like this (this code is untested):
# Filter to the desired attributes
sum_yz = df[df['attribute'].isin(['y', 'z'])]
# Set the new 'attribute' value
sum_yz['attribute'] = 'sum_yz'
# Group by and sum
sum_yz = sum_yz.groupby(['prod', 'attribute']).sum().reset_index()
# Add it the end of the data frame
df = pd.concat([df, sum_yz])
You can use df.groupby()
and then combine the groupby-outcome with the original df
# Create groupby DataFrame
df_grp = df[df['attribute'].isin(['y', 'z'])].groupby(['prod']).sum()
df_grp.reset_index(inplace=True)
df_grp['attribute'] = 'sum_yz'
# Combine with original dataframe
df = pd.concat([df, df_grp])
One idea with dictionaries, but slowier if large DataFrame:
def f(x):
d = x[x['attribute'].isin(['y','z'])].sum()
d1 = {'prod': x.name, 'attribute':'sum_yz'}
x = x.append({**d, **d1},ignore_index=True)
return x
df = df.groupby('prod', sort=False).apply(f).reset_index(drop=True)
print (df)
prod attribute number1 number2 number3
0 a x 1 10 1
1 a y 2 2 4
2 a z 2 3 3
3 a sum_yz 4 5 7
4 b x 3 3 5
5 b y 4 1 7
6 b z 3 2 1
7 b sum_yz 7 3 8
8 c x 5 3 3
9 c y 1 1 0
10 c z 1 1 1
11 c sum_yz 2 2 1
Or if possible sorting values of product first filter by Series.isin
, aggregate sum
, add to original with replace NaN
by DataFrame.fillna
and last sorting by DataFrame.sort_values
with ignore_index
for default index:
df = (df.append(df[df['attribute'].isin(['y', 'z'])]
.groupby('prod', as_index=False)
.sum()
).fillna({'attribute': 'sum_yz'})
.sort_values('prod', ignore_index=True))
print (df)
prod attribute number1 number2 number3
0 a x 1 10 1
1 a y 2 2 4
2 a z 2 3 3
3 a sum_yz 4 5 7
4 b x 3 3 5
5 b y 4 1 7
6 b z 3 2 1
7 b sum_yz 7 3 8
8 c x 5 3 3
9 c y 1 1 0
10 c z 1 1 1
11 c sum_yz 2 2 1
You can use pandas concat after the groupby:
result = df.groupby(["prod", df.attribute.isin(["y", "z"])]).sum().loc[:, True, :]
result = result.reset_index()
result.insert(1, "attribute", "sum_yz")
pd.concat([df, result]).sort_values("prod", ignore_index=True)
prod attribute number1 number2 number3
0 a x 1 10 1
1 a y 2 2 4
2 a z 2 3 3
3 a sum_yz 4 5 7
4 b x 3 3 5
5 b y 4 1 7
6 b z 3 2 1
7 b sum_yz 7 3 8
8 c x 5 3 3
9 c y 1 1 0
10 c z 1 1 1
11 c sum_yz 2 2 1
this is simple and works fine
dr=df[df['attribute']!='x'].groupby('prod').sum().reset_index()
dr['attribute']='sum_yz'
result=pd.concat([df,dr]).sort_values('prod')
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.