I'm working on a project where my original dataframe is:
A B C label
0 1 2 2 Nan
1 2 4 5 7
2 3 6 5 Nan
3 4 8 7 Nan
4 5 10 3 8
5 6 12 4 8
But, I have an array with new labels for certain points (for that I only used columns A and B) in the original dataframe. Something like this:
X_labeled = [[2, 4], [3,6]]
y_labeled = [5,9]
My goal is to add the new labels to the original dataframe. I know that the combination of A and B unique is. What is the fastest way to assign the new label to the correct row?
This is my try:
y_labeled = np.array(y).astype('float64')
current_position = 0
for point in X_labeled:
row = df.loc[(df['A'] == point[0]) & (df['B'] == point[1])]
df.at[row.index, 'label'] = y_labeled[current_position]
current_position += 1
Wanted output (rows with index 1 and 2 are changed):
A B C label
0 1 2 2 Nan
1 2 4 5 5
2 3 6 5 9
3 4 8 7 Nan
4 5 10 3 8
5 6 12 4 8
For small datasets may this be okay with I'm currently using it for datasets with more than 25000 labels. Is there a way that is faster?
Also, in some cases I used all columns expect the column 'label'. That dataframe exists out of 64 columns so my method can not be used here. Has someone an idea to improve this?
Thanks in advance
Best solution is to make your arrays into a dataframe and use df.update():
new = pd.DataFrame(X_labeled, columns=['A', 'B'])
new['label'] = y_labeled
new = new.set_index(['A', 'B'])
df = df.set_index(['A', 'B'])
df.update(new)
df = df.reset_index()
Here's a numpy based approach aimed at performance. To vectorize this we want a way to check membership of the rows in X_labeled in columns A and B. So what we can do, is view these two columns as 1D arrays (based on this answer) and then we can use np.in1d to index the dataframe and assign the values in y_labeled:
import numpy as np
X_labeled = [[2, 4], [3,6]]
y_labeled = [5,9]
a = df.values[:,:2].astype(int) #indexing on A and B
def view_as_1d(a):
a = np.ascontiguousarray(a)
return a.view(np.dtype((np.void, a.dtype.itemsize * a.shape[-1])))
ix = np.in1d(view_as_1d(a), view_as_1d(X_labeled))
df.loc[ix, 'label'] = y_labeled
print(df)
A B C label
0 1 2 2 Nan
1 2 4 5 5
2 3 6 5 9
3 4 8 7 Nan
4 5 10 3 8
5 6 12 4 8
Related
I have a dataframe:
t = pd.Series([2,4,6,8,10,12],index= index)
df1 = pd.DataFrame(s,columns = ["MUL1"])
df1["MUL2"] =t
MUL1 MUL2
0 1 2
1 2 4
2 2 6
3 3 8
4 3 10
5 6 12
and another dataframe:
u = pd.Series([1,2,3,6],index= index)
v = pd.Series([2,8,10,12],index= index)
df2 = pd.DataFrame(u,columns = ["MUL3"])
df2["MUL4"] =v
Now I want a new dataframe which looks like the following:
MUL6 MUL7
0 1 2
1 2 8
2 2 8
3 3 10
4 3 10
5 6 12
By combining the first 2 dataframes.
I have tried the following:
X1 = df1.to_numpy()
X2 = df2.to_numpy()
list = []
for i in range(X1.shape[0]):
for j in range(X2.shape[0]):
if X1[i, -1] == X2[j, -1]:
list.append(X2[X1[i, -1]==X2[j, -1], -1])
I was trying to convert the dataframes to numpy arrays so I can iterate through them to get a new array that I can convert back to a dataframe. But the size of the new dataframe is not equal to size of the first dataframe. Please I would appreciate any help. Thanks.
Although the details of the logic are cryptic, I believe that you want a merge:
(df1[['MUL1']].rename(columns={'MUL1': 'MUL6'})
.merge(df2.rename(columns={'MUL3': 'MUL6', 'MUL4': 'MUL7'}),
on='MUL6', how='left')
)
output:
MUL6 MUL7
0 1 2
1 2 8
2 2 8
3 3 10
4 3 10
5 6 12
I tried to create a data frame df using the below code :
import numpy as np
import pandas as pd
index = [0,1,2,3,4,5]
s = pd.Series([1,2,3,4,5,6],index= index)
t = pd.Series([2,4,6,8,10,12],index= index)
df = pd.DataFrame(s,columns = ["MUL1"])
df["MUL2"] =t
print df
MUL1 MUL2
0 1 2
1 2 4
2 3 6
3 4 8
4 5 10
5 6 12
While trying to create the same data frame using the below syntax, I am getting a wierd output.
df = pd.DataFrame([s,t],columns = ["MUL1","MUL2"])
print df
MUL1 MUL2
0 NaN NaN
1 NaN NaN
Please explain why the NaN is being displayed in the dataframe when both the Series are non empty and why only two rows are getting displayed and no the rest.
Also provide the correct way to create the data frame same as has been mentioned above by using the columns argument in the pandas DataFrame method.
One of the correct ways would be to stack the array data from the input list holding those series into columns -
In [161]: pd.DataFrame(np.c_[s,t],columns = ["MUL1","MUL2"])
Out[161]:
MUL1 MUL2
0 1 2
1 2 4
2 3 6
3 4 8
4 5 10
5 6 12
Behind the scenes, the stacking creates a 2D array, which is then converted to a dataframe. Here's what the stacked array looks like -
In [162]: np.c_[s,t]
Out[162]:
array([[ 1, 2],
[ 2, 4],
[ 3, 6],
[ 4, 8],
[ 5, 10],
[ 6, 12]])
If remove columns argument get:
df = pd.DataFrame([s,t])
print (df)
0 1 2 3 4 5
0 1 2 3 4 5 6
1 2 4 6 8 10 12
Then define columns - if columns not exist get NaNs column:
df = pd.DataFrame([s,t], columns=[0,'MUL2'])
print (df)
0 MUL2
0 1.0 NaN
1 2.0 NaN
Better is use dictionary:
df = pd.DataFrame({'MUL1':s,'MUL2':t})
print (df)
MUL1 MUL2
0 1 2
1 2 4
2 3 6
3 4 8
4 5 10
5 6 12
And if need change columns order add columns parameter:
df = pd.DataFrame({'MUL1':s,'MUL2':t}, columns=['MUL2','MUL1'])
print (df)
MUL2 MUL1
0 2 1
1 4 2
2 6 3
3 8 4
4 10 5
5 12 6
More information is in dataframe documentation.
Another solution by concat - DataFrame constructor is not necessary:
df = pd.concat([s,t], axis=1, keys=['MUL1','MUL2'])
print (df)
MUL1 MUL2
0 1 2
1 2 4
2 3 6
3 4 8
4 5 10
5 6 12
A pandas.DataFrame takes in the parameter data that can be of type ndarray, iterable, dict, or dataframe.
If you pass in a list it will assume each member is a row. Example:
a = [1,2,3]
b = [2,4,6]
df = pd.DataFrame([a, b], columns = ["Col1","Col2", "Col3"])
# output 1:
Col1 Col2 Col3
0 1 2 3
1 2 4 6
You are getting NaN because it expects index = [0,1] but you are giving [0,1,2,3,4,5]
To get the shape you want, first transpose the data:
data = np.array([a, b]).transpose()
How to create a pandas dataframe
import pandas as pd
a = [1,2,3]
b = [2,4,6]
df = pd.DataFrame(dict(Col1=a, Col2=b))
Output:
Col1 Col2
0 1 2
1 2 4
2 3 6
I am using pandas v0.25.3. and am inexperienced but learning.
I have a dataframe and would like to swap the contents of two columns leaving the columns labels and sequence intact.
df = pd.DataFrame ({"A": [(1),(2),(3),(4)],
'B': [(5),(6),(7),(8)],
'C': [(9),(10),(11),(12)]})
This yields a dataframe,
A B C
0 1 5 9
1 2 6 10
2 3 7 11
3 4 8 12
I want to swap column contents B and C to get
A B C
0 1 9 5
1 2 10 6
2 3 11 7
3 4 12 8
I have tried looking at pd.DataFrame.values which sent me to numpy array and advanced slicing and got lost.
Whats the simplest way to do this?.
You can assign numpy array:
#pandas 0.24+
df[['B','C']] = df[['C','B']].to_numpy()
#oldier pandas versions
df[['B','C']] = df[['C','B']].values
Or use DataFrame.assign:
df = df.assign(B = df.C, C = df.B)
print (df)
A B C
0 1 9 5
1 2 10 6
2 3 11 7
3 4 12 8
Or just use:
df['B'], df['C'] = df['C'], df['B'].copy()
print(df)
Output:
A B C
0 1 9 5
1 2 10 6
2 3 11 7
3 4 12 8
You can also swap the labels:
df.columns = ['A','C','B']
If your DataFrame is very large, I believe this would require less from your computer than copying all the data.
If the order of the columns is important, you can then reorder them:
df = df.reindex(['A','B','C'], axis=1)
This sounds a bit weird, but I think that's exactly what I needed now:
I got several pandas dataframes that contains columns with float numbers, for example:
a b c
0 0 1 2
1 3 4 5
2 6 7 8
Now I want to add a column, with only one row, and the value is equal to the average of column 'a', in this case, is 3.0. So the new dataframe will looks like this:
a b c average
0 0 1 2 3.0
1 3 4 5
2 6 7 8
And all the rows below are empty.
I've tried things like df['average'] = np.mean(df['a']) but that give me a whole column of 3.0. Any help will be appreciated.
Assign a series, this is cleaner.
df['average'] = pd.Series(df['a'].mean(), index=df.index[[0]])
Or, even better, assign with loc:
df.loc[df.index[0], 'average'] = df['a'].mean().item()
Filling NaNs is straightforward, you can do
df['average'] = df['average'].fillna('')
df
a b c average
0 0 1 2 3
1 3 4 5
2 6 7 8
Can do something like:
df['average'] = [np.mean(df['a'])]+['']*(len(df)-1)
Here is a full example:
import pandas as pd
import numpy as np
df = pd.DataFrame(
[(0,1,2), (3,4,5), (6,7,8)],
columns=['a', 'b', 'c'])
print(df)
a b c
0 0 1 2
1 3 4 5
2 6 7 8
df['average'] = ''
df['average'][0] = df['a'].mean()
print(df)
a b c average
0 0 1 2 3
1 3 4 5
2 6 7 8
I am currently using Pandas and Python to handle much of the repetitive tasks, I need done for my master thesis. At this point, I have written some code (with help from stack overflow) that, based on some event dates in one file, finds a start and end date to use as a date range in another file. These dates are then located and appended to an empty list, which I can then output to excel. However, using the below code I get a dataframe with 5 columns and 400.000 + rows (which is basically what I want), but not how I want the data outputted to excel. Below is my code:
end_date = pd.DataFrame(data=(df_sample['Date']-pd.DateOffset(days=2)))
start_date = pd.DataFrame(data=(df_sample['Date']-pd.offsets.BDay(n=252)))
merged_dates = pd.merge(end_date,start_date,left_index=True,right_index=True)
ff_factors = []
for index, row in merged_dates.iterrows():
time_range= (df['Date'] > row['Date_y']) & (df['Date'] <= row['Date_x'])
df_factor = df.loc[time_range]
ff_factors.append(df_factor)
appended_data = pd.concat(ff_factors, axis=0)
I need the data to be 5 columns and 250 rows (columns are variable identifiers) side by side, so that when outputting it to excel I have, for example column A-D and then 250 rows for each column. This then needs to be repeated for column E-H and so on. Using iloc, I can locate the 250 observations using appended_data.iloc[0:250], with both 5 columns and 250 rows, and then output it to excel.
Are the any way for me to automate the process, so that after selecting the first 250 and outputting it to excel, it selects the next 250 and outputs it next to the first 250 and so on?
I hope the above is precise and clear, else I'm happy to elaborate!
EDIT:
The above picture illustrate what I get when outputting to excel; 5 columns and 407.764 rows. What I needed is to get this split up into the following way:
The second picture illustrates how I needed the total sample to be split up. The first five columns and corresponding 250 rows needs to be as the second picture. When I do the next split using iloc[250:500], I will get the next 250 rows, which needs to be added after the initial five columns and so on.
You can do this with a combination of np.reshape, which can be made to behave as desired on individual columns, and which should be much faster than a loop through the rows, and pd.concat, to join the dataframes it makes back together:
def reshape_appended(df, target_rows, pad=4):
df = df.copy() # don't modify in-place
# below line adds strings, '0000',...,'0004' to the column names
# this ensures sorting the columns preserves the order
df.columns = [str(i).zfill(pad)+df.columns[i] for i in range(len(df.columns))]
#target number of new columns per column in df
target_cols = len(df.index)//target_rows
last_group = pd.DataFrame()
# below conditional fires if there will be leftover rows - % is mod
if len(df.index)%target_rows != 0:
last_group = df.iloc[-(len(df.index)%target_rows):].reset_index(drop=True)
df = df.iloc[:-(len(df.index)%target_rows)] # keep rows that divide nicely
#this is a large list comprehension, that I'll elaborate on below
groups = [pd.DataFrame(df[col].values.reshape((target_rows, target_cols),
order='F'),
columns=[str(i).zfill(pad)+col for i in range(target_cols)])
for col in df.columns]
if not last_group.empty: # if there are leftover rows, add them back
last_group.columns = [pad*'9'+col for col in last_group.columns]
groups.append(last_group)
out = pd.concat(groups, axis=1).sort_index(axis=1)
out.columns = out.columns.str[2*pad:] # remove the extra characters in the column names
return out
last_group takes care of any rows that don't divide evenly into sets of 250. The playing around with column names enforces proper sorting order.
df[col].values.reshape((target_rows, target_cols), order='F')
Reshapes the values in the column col of df into the shape specified by the tuple (target_rows, target_cols), with the ordering Fortran uses, indicated by F.
columns=[str(i).zfill(pad)+col for i in range(target_cols)]
is just giving names to these columns, with any eye to establishing proper ordering afterward.
Ex:
df = pd.DataFrame(np.random.randint(0, 10, (23, 3)), columns=list('abc'))
reshape_appended(df, 5)
Out[160]:
a b c a b c a b c a b c a b c
0 8 3 0 4 1 9 5 4 7 2 3 4 5.0 7.0 2.0
1 1 6 1 3 5 1 1 6 0 5 9 4 6.0 0.0 1.0
2 3 1 3 4 3 8 9 3 9 8 7 8 7.0 3.0 2.0
3 4 0 1 5 5 6 6 4 4 0 0 3 NaN NaN NaN
4 9 7 3 5 7 4 6 5 8 9 5 5 NaN NaN NaN
df
Out[161]:
a b c
0 8 3 0
1 1 6 1
2 3 1 3
3 4 0 1
4 9 7 3
5 4 1 9
6 3 5 1
7 4 3 8
8 5 5 6
9 5 7 4
10 5 4 7
11 1 6 0
12 9 3 9
13 6 4 4
14 6 5 8
15 2 3 4
16 5 9 4
17 8 7 8
18 0 0 3
19 9 5 5
20 5 7 2
21 6 0 1
22 7 3 2
My best guess to solving the problem would be to try and loop, until the counter is greater than length, so
i = 250 # counter
j = 0 # left limit
for x in range(len("your dataframe")):
appended_data.iloc[j:i]
i+=250
if i > len("your df"):
appended_data.iloc[j:(len("your df"))
break
else:
j = i