问题
I have two pandas dataframes df1 (of length 2) and df2 (of length about 30 rows). Index values of df1 are always different and never occur in df2. I would like to add the average of columns from df1 to corresponding columns of df2. Example: add 0.6 to all rows of c1 and 0.9 to all rows of c2 etc ...
df1:
Date c1 c2 c3 c4 c5 c6 ... c10
2017-09-10 0.5 0.6 1.2 0.7 1.3 1.8 ... 1.3
2017-09-11 0.7 1.2 1.3 0.4 0.7 0.4 ... 1.5
df2:
Date c1 c2 c3 c4 c5 c6 ... c10
2017-09-12 0.9 0.1 1.4 0.9 1.5 1.9 ... 1.9
2017-09-13 0.2 1.8 1.2 1.4 2.7 0.8 ... 1.1
: :
: :
2017-10-10 1.5 0.9 1.5 0.9 1.6 1.8 ... 1.7
2017-10-11 2.7 1.1 1.9 0.4 0.8 0.8 ... 1.3
How can I do that ?
回答1:
When using mean on df1, it calculates over each column by default and produces a pd.Series.
When adding adding a pd.Series to a pd.DataFrame it aligns the index of the pd.Series with the columns of the pd.DataFrame and broadcasts along the index of the pd.DataFrame... by default.
The only tricky bit is handling the Date column.
Option 1
m = df1.mean()
df2.loc[:, m.index] += m
df2
Date c1 c2 c3 c4 c5 c6 c10
0 2017-09-12 1.5 1.0 2.65 1.45 2.5 3.0 3.3
1 2017-09-13 0.8 2.7 2.45 1.95 3.7 1.9 2.5
2 2017-10-10 2.1 1.8 2.75 1.45 2.6 2.9 3.1
3 2017-10-11 3.3 2.0 3.15 0.95 1.8 1.9 2.7
If I know that 'Date' is always in the first column, I can:
df2.iloc[:, 1:] += df1.mean()
df2
Date c1 c2 c3 c4 c5 c6 c10
0 2017-09-12 1.5 1.0 2.65 1.45 2.5 3.0 3.3
1 2017-09-13 0.8 2.7 2.45 1.95 3.7 1.9 2.5
2 2017-10-10 2.1 1.8 2.75 1.45 2.6 2.9 3.1
3 2017-10-11 3.3 2.0 3.15 0.95 1.8 1.9 2.7
Option 2
Notice that I use the append=True parameter in the set_index just incase there are things in the index you don't want to mess up.
df2.set_index('Date', append=True).add(df1.mean()).reset_index('Date')
Date c1 c2 c3 c4 c5 c6 c10
0 2017-09-12 1.5 1.0 2.65 1.45 2.5 3.0 3.3
1 2017-09-13 0.8 2.7 2.45 1.95 3.7 1.9 2.5
2 2017-10-10 2.1 1.8 2.75 1.45 2.6 2.9 3.1
3 2017-10-11 3.3 2.0 3.15 0.95 1.8 1.9 2.7
If you don't care about the index, you can shorten this to
df2.set_index('Date').add(df1.mean()).reset_index()
Date c1 c2 c3 c4 c5 c6 c10
0 2017-09-12 1.5 1.0 2.65 1.45 2.5 3.0 3.3
1 2017-09-13 0.8 2.7 2.45 1.95 3.7 1.9 2.5
2 2017-10-10 2.1 1.8 2.75 1.45 2.6 2.9 3.1
3 2017-10-11 3.3 2.0 3.15 0.95 1.8 1.9 2.7
回答2:
If all columns are in both data frames, then just
for col in df2.columns:
df2[col] = df2[col] + df1[col].mean()
if the columns are not necessarily in both then:
for col in df2.columns:
if col in df1.columns:
df2[col] = df2[col] + df1[col].mean()
回答3:
There is probably a more efficient way but here is a quick and dirty solution. I hope this helps!
d = {'c1': [0.5,0.7], 'c2': [0.6,1.2],'c3': [1.2,1.3]}
df1 = pd.DataFrame(data=d, index=['2017-09-10','2017-09-11'])
df2 = pd.DataFrame(data=d, index=['2017-09-12','2017-09-13'])
df1
Date c1 c2 c3
2017-09-10 0.5 0.6 1.2
2017-09-11 0.7 1.2 1.3
df2
Date c1 c2 c3
2017-09-12 0.5 0.6 1.2
2017-09-13 0.7 1.2 1.3
The averages of each column in df1 can be obtained using the describe() function
df1.describe().ix['mean']
c1 0.60
c2 0.90
c3 1.25
And now, simply add the series to df2
df2 + df1.describe().ix['mean']
Date c1 c2 c3
2017-09-12 1.1 1.5 2.45
2017-09-13 1.3 2.1 2.55
回答4:
This could be another way of doing it , just simplified this a little bit
import pandas as pd
import numpy as np
from datetime import datetime, timedelta
date_today=datetime.now()
#Creating df1 & df2
df1=pd.DataFrame(
{
'Date':[date_today,date_today],
'c1':[0.5,0.4],
'c2':[0.6,0.3]
}
)
df2=pd.DataFrame(
{
'Date':[date_today,date_today,date_today],
'c1':[0.9,0.7,0.6],
'c2':[0.8,0.4,0.3]
}
)
#getting average of column c1
avg=df1["c1"].mean()
#Adding the average to your existing column of df2
df2['c1']+avg
来源：https://stackoverflow.com/questions/46967581/adding-values-to-all-rows-of-dataframe