I'd like to get the number of month in between these dates (between max and minimum date) and keep the same order in the groupby
One of possible solutions is to start from a datesac - the result
of your grouping (presented in your picture).
I also assume that ORDER_INST column of your source DataFrame is of datetime type (not a string) and hence just this type has also level 1 of
the MultiIndex in datesac.
To compute the month span separately for each MRN (level 0 of the
MultiIndex), define a function, to be applied to each group:
def monthSpan(grp):
dates = grp.index.get_level_values(1)
return (dates.max().to_period('M') - dates.min().to_period('M')).n
Then add MonthSpan column to your df, running:
datesac['MonthSpan'] = datesac.groupby(level=0).transform(monthSpan);
The result is:
List MonthSpan
MRN ORDER_INST
1000031 2010-04-12 0 11
2010-04-16 0 11
2010-04-17 0 11
2010-04-18 0 11
2011-03-01 0 11
9017307 2018-11-27 0 7
2019-02-04 0 7
2019-04-25 0 7
2019-05-14 0 7
2019-06-09 0 7
Pandas does not allow item assignments to a groupby object (a new column cannot be added to a groupby object) so the operation will have to be split. One solution is first calculate the month difference from the groupby object, merge the dataframes together, and then groupby again.
Create the first groupby object:
datesac = acdates.groupby(['MRN'])
Calculate the difference in months between each group and join to the original dataframe (or a new dataframe). This method requires numpy so import as necessary
import numpy as np
acdates_new = pd.merge(
left=acdates,
right=((datesac['ORDER_INST'].max() - df_group['ORDER_INST'].min())/np.timedelta64(1, 'M')).astype('int').rename("DATE_DIFF"),
left_on='MRN',
right_index=True
)
Regroup
datesac = acdates_new.groupby(['MRN'])
Related
I have the following data frame in IPython, where each row is a single stock:
In [261]: bdata
Out[261]:
<class 'pandas.core.frame.DataFrame'>
Int64Index: 21210 entries, 0 to 21209
Data columns:
BloombergTicker 21206 non-null values
Company 21210 non-null values
Country 21210 non-null values
MarketCap 21210 non-null values
PriceReturn 21210 non-null values
SEDOL 21210 non-null values
yearmonth 21210 non-null values
dtypes: float64(2), int64(1), object(4)
I want to apply a groupby operation that computes cap-weighted average return across everything, per each date in the "yearmonth" column.
This works as expected:
In [262]: bdata.groupby("yearmonth").apply(lambda x: (x["PriceReturn"]*x["MarketCap"]/x["MarketCap"].sum()).sum())
Out[262]:
yearmonth
201204 -0.109444
201205 -0.290546
But then I want to sort of "broadcast" these values back to the indices in the original data frame, and save them as constant columns where the dates match.
In [263]: dateGrps = bdata.groupby("yearmonth")
In [264]: dateGrps["MarketReturn"] = dateGrps.apply(lambda x: (x["PriceReturn"]*x["MarketCap"]/x["MarketCap"].sum()).sum())
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
/mnt/bos-devrnd04/usr6/home/espears/ws/Research/Projects/python-util/src/util/<ipython-input-264-4a68c8782426> in <module>()
----> 1 dateGrps["MarketReturn"] = dateGrps.apply(lambda x: (x["PriceReturn"]*x["MarketCap"]/x["MarketCap"].sum()).sum())
TypeError: 'DataFrameGroupBy' object does not support item assignment
I realize this naive assignment should not work. But what is the "right" Pandas idiom for assigning the result of a groupby operation into a new column on the parent dataframe?
In the end, I want a column called "MarketReturn" than will be a repeated constant value for all indices that have matching date with the output of the groupby operation.
One hack to achieve this would be the following:
marketRetsByDate = dateGrps.apply(lambda x: (x["PriceReturn"]*x["MarketCap"]/x["MarketCap"].sum()).sum())
bdata["MarketReturn"] = np.repeat(np.NaN, len(bdata))
for elem in marketRetsByDate.index.values:
bdata["MarketReturn"][bdata["yearmonth"]==elem] = marketRetsByDate.ix[elem]
But this is slow, bad, and unPythonic.
In [97]: df = pandas.DataFrame({'month': np.random.randint(0,11, 100), 'A': np.random.randn(100), 'B': np.random.randn(100)})
In [98]: df.join(df.groupby('month')['A'].sum(), on='month', rsuffix='_r')
Out[98]:
A B month A_r
0 -0.040710 0.182269 0 -0.331816
1 -0.004867 0.642243 1 2.448232
2 -0.162191 0.442338 4 2.045909
3 -0.979875 1.367018 5 -2.736399
4 -1.126198 0.338946 5 -2.736399
5 -0.992209 -1.343258 1 2.448232
6 -1.450310 0.021290 0 -0.331816
7 -0.675345 -1.359915 9 2.722156
While I'm still exploring all of the incredibly smart ways that apply concatenates the pieces it's given, here's another way to add a new column in the parent after a groupby operation.
In [236]: df
Out[236]:
yearmonth return
0 201202 0.922132
1 201202 0.220270
2 201202 0.228856
3 201203 0.277170
4 201203 0.747347
In [237]: def add_mkt_return(grp):
.....: grp['mkt_return'] = grp['return'].sum()
.....: return grp
.....:
In [238]: df.groupby('yearmonth').apply(add_mkt_return)
Out[238]:
yearmonth return mkt_return
0 201202 0.922132 1.371258
1 201202 0.220270 1.371258
2 201202 0.228856 1.371258
3 201203 0.277170 1.024516
4 201203 0.747347 1.024516
As a general rule when using groupby(), if you use the .transform() function pandas will return a table with the same length as your original. When you use other functions like .sum() or .first() then pandas will return a table where each row is a group.
I'm not sure how this works with apply but implementing elaborate lambda functions with transform can be fairly tricky so the strategy that I find most helpful is to create the variables I need, place them in the original dataset and then do my operations there.
If I understand what you're trying to do correctly first you can calculate the total market cap for each group:
bdata['group_MarketCap'] = bdata.groupby('yearmonth')['MarketCap'].transform('sum')
This will add a column called "group_MarketCap" to your original data which would contain the sum of market caps for each group. Then you can calculate the weighted values directly:
bdata['weighted_P'] = bdata['PriceReturn'] * (bdata['MarketCap']/bdata['group_MarketCap'])
And finally you would calculate the weighted average for each group using the same transform function:
bdata['MarketReturn'] = bdata.groupby('yearmonth')['weighted_P'].transform('sum')
I tend to build my variables this way. Sometimes you can pull off putting it all in a single command but that doesn't always work with groupby() because most of the time pandas needs to instantiate the new object to operate on it at the full dataset scale (i.e. you can't add two columns together if one doesn't exist yet).
Hope this helps :)
May I suggest the transform method (instead of aggregate)? If you use it in your original example it should do what you want (the broadcasting).
I did not find a way to make assignment to the original dataframe. So I just store the results from the groups and concatenate them. Then we sort the concatenated dataframe by index to get the original order as the input dataframe. Here is a sample code:
In [10]: df = pd.DataFrame({'month': np.random.randint(0,11, 100), 'A': np.random.randn(100), 'B': np.random.randn(100)})
In [11]: df.head()
Out[11]:
month A B
0 4 -0.029106 -0.904648
1 2 -2.724073 0.492751
2 7 0.732403 0.689530
3 2 0.487685 -1.017337
4 1 1.160858 -0.025232
In [12]: res = []
In [13]: for month, group in df.groupby('month'):
...: new_df = pd.DataFrame({
...: 'A^2+B': group.A ** 2 + group.B,
...: 'A+B^2': group.A + group.B**2
...: })
...: res.append(new_df)
...:
In [14]: res = pd.concat(res).sort_index()
In [15]: res.head()
Out[15]:
A^2+B A+B^2
0 -0.903801 0.789282
1 7.913327 -2.481270
2 1.225944 1.207855
3 -0.779501 1.522660
4 1.322360 1.161495
This method is pretty fast and extensible. You can derive any feature here.
Note: If the dataframe is too large, concat may cause you MMO error.
I have a function which sends automated messages to clients, and takes as input all the columns from a dataframe like the one below.
name
phone
status
date
name_1
phone_1
sending
today
name_2
phone_2
sending
yesterday
I iterate through the dataframe with a pandas apply (axis=1) and use the values on the columns of each row as inputs to my function. At the end of it, after sending, it changes the status to "sent". The thing is I only want to send to the clients whose date reference is "today". Now, with pandas.apply(axis=1) this is perfectly doable, but in order to slice the clients with "today" value, I need to:
create a new dataframe with today's value,
remove it from the original, and then
reappend it to the original.
I thought about running through the whole dataframe and ignore the rows which have dates different than "today", but if my dataframe keeps growing, I'm afraid of the whole process becoming slower.
I saw examples of this being done with mask, although usually people only use 1 column, and I need more than just the one. Is there any way to do this with pandas apply?
Thank you.
I think you can use .loc to filter the data and apply func to it.
In [13]: df = pd.DataFrame(np.random.rand(5,5))
In [14]: df
Out[14]:
0 1 2 3 4
0 0.085870 0.013683 0.221890 0.533393 0.622122
1 0.191646 0.331533 0.259235 0.847078 0.649680
2 0.334781 0.521263 0.402030 0.973504 0.903314
3 0.189793 0.251130 0.983956 0.536816 0.703726
4 0.902107 0.226398 0.596697 0.489761 0.535270
if we want double the values of rows where the value in first column > 0.3
Out[16]:
0 1 2 3 4
2 0.334781 0.521263 0.402030 0.973504 0.903314
4 0.902107 0.226398 0.596697 0.489761 0.535270
In [18]: df.loc[df[0] > 0.3] = df.loc[df[0] > 0.3].apply(lambda x: x*2, axis=1)
In [19]: df
Out[19]:
0 1 2 3 4
0 0.085870 0.013683 0.221890 0.533393 0.622122
1 0.191646 0.331533 0.259235 0.847078 0.649680
2 0.669563 1.042527 0.804061 1.947008 1.806628
3 0.189793 0.251130 0.983956 0.536816 0.703726
4 1.804213 0.452797 1.193394 0.979522 1.070540
I'm creating a Pandas dataframe from an existing file and it ends up essentially like this.
import pandas as pd
import datetime
data = [[i, i+1] for i in range(14)]
index = pd.date_range(start=datetime.date(2019,1,1), end=datetime.date(2020,2,1), freq='MS')
columns = ['col1', 'col2']
df = pd.DataFrame(data, index, columns)
Notice that this doesn't go all the way up to the present -- often the file I'm pulling from is a month or two behind. What I then need to do is add on any missing months and fill them with the same value as the previous year.
So in this case I need to add another row that is
2020-03-01 2 3
It could be anywhere from 0-2 rows that need to be added to the end of the dataframe at a given point in time. What's the best way to do this?
Note: The data here is not real so please don't take advantage of the simple pattern of entries I gave above. It was just a quick way to fill two columns of a table as an example.
If I understand your problem, then the following should help you. This does assume that you always have data 12 months ago however. You can define a new DataFrame which includes the months up to the most recent date.
# First create the new index. Get the most recent date and add an offset.
start, end = df.index[-1] + pd.DateOffset(), pd.Timestamp.now()
index_new = pd.date_range(start, end, freq='MS')
Create your DataFrame
# Get the data from the previous year.
data = df.loc[index_new - pd.DateOffset(years=1)].values
df_new = pd.DataFrame(data, index = index_new, columns=df.columns)
which looks like
col1 col2
2020-03-01 2 3
then just use;
pd.concat([df, df_new], axis=0)
Which gives
col1 col2
2019-01-01 0 1
2019-02-01 1 2
2019-03-01 2 3
... ... ...
2020-02-01 13 14
2020-03-01 2 3
Note
This also works for cases where the number of months missing is greater than 1.
Edit
Slightly different variation
# Create series with missing months added.
# Get the corresponding data 12 months prior.
s = pd.date_range(df.index[0], pd.Timestamp.now(), freq='MS')
fill = df.loc[s[~s.isin(df.index)] - pd.DateOffset(years=1)]
# Reindex the original dataframe
df = df.reindex(s)
# Find the dates to fill and replace with lagged data
df.iloc[-1 * fill.shape[0]:] = fill.values
Initial Note
I already got this running, but it takes a very long time to execute. My DataFrame is around 500MB large. I am hoping to hear some feedback on how to execute this as quickly as possible.
Problem Statement
I want to normalize the DataFrame columns by the mean of the column's values during each month. An added complexity is that I have a column named group which denotes a different sensor in which the parameter (column) was measured. Therefore, the analysis needs to iterate around group and each month.
DF example
X Y Z group
2019-02-01 09:30:07 1 2 1 'grp1'
2019-02-01 09:30:23 2 4 3 'grp2'
2019-02-01 09:30:38 3 6 5 'grp1'
...
Code (Functional, but slow)
This is the code that I used. Coding annotations provide descriptions of most lines. I recognize that the three for loops are causing this runtime issue, but I do not have the foresight to see a way around it. Does anyone know any
# Get mean monthly values for each group
mean_per_month_unit = process_df.groupby('group').resample('M', how='mean')
# Store the monthly dates created in last line into a list called month_dates
month_dates = mean_per_month_unit.index.get_level_values(1)
# Place date on multiIndex columns. future note: use df[DATE, COL_NAME][UNIT] to access mean value
mean_per_month_unit = mean_per_month_unit.unstack().swaplevel(0,1,1).sort_index(axis=1)
divide_df = pd.DataFrame().reindex_like(df)
process_cols.remove('group')
for grp in group_list:
print(grp)
# Iterate through month
for mnth in month_dates:
# Make mask where month and group
mask = (df.index.month == mnth.month) & (df['group'] == grp)
for col in process_cols:
# Set values of divide_df
divide_df.iloc[mask.tolist(), divide_df.columns.get_loc(col)] = mean_per_month_unit[mnth, col][grp]
# Divide process_df with divide_df
final_df = process_df / divide_df.values
EDIT: Example data
Here is the data in CSV format.
EDIT2: Current code (according to current answer)
def normalize_df(df):
df['month'] = df.index.month
print(df['month'])
df['year'] = df.index.year
print(df['year'])
def find_norm(x, df_col_list): # x is a row in dataframe, col_list is the list of columns to normalize
agg = df.groupby(by=['group', 'month', 'year'], as_index=True).mean()
print("###################", x.name, x['month'])
for column in df_col_list: # iterate over col list, find mean from aggregations, and divide the value by
print(column)
mean_col = agg.loc[(x['group'], x['month'], x['year']), column]
print(mean_col)
col_name = "norm" + str(column)
x[col_name] = x[column] / mean_col # norm
return x
normalize_cols = df.columns.tolist()
normalize_cols.remove('group')
#normalize_cols.remove('mode')
df2 = df.apply(find_norm, df_col_list = normalize_cols, axis=1)
The code runs perfectly for one iteration and then it fails with the error:
KeyError: ('month', 'occurred at index 2019-02-01 11:30:17')
As I said, it runs correctly once. However, it iterates over the same row again and then fails. I see according to df.apply() documentation that the first row always runs twice. I'm just not sure why this fails on the second time through.
Assuming that the requirement is to group the columns by mean and the month, here is another approach:
Create new columns - month and year from the index. df.index.month can be used for this provided the index is of type DatetimeIndex
type(df.index) # df is the original dataframe
#pandas.core.indexes.datetimes.DatetimeIndex
df['month'] = df.index.month
df['year'] = df.index.year # added year assuming the grouping occurs per grp per month per year. No need to add this column if year is not to be considered.
Now, group over (grp, month, year) and aggregate to find mean of every column. (Added year assuming the grouping occurs per grp per month per year. No need to add this column if year is not to be considered.)
agg = df.groupby(by=['grp', 'month', 'year'], as_index=True).mean()
Use a function to calculate the normalized values and use apply() over the original dataframe
def find_norm(x, df_col_list): # x is a row in dataframe, col_list is the list of columns to normalize
for column in df_col_list: # iterate over col list, find mean from aggregations, and divide the value by the mean.
mean_col = agg.loc[(str(x['grp']), x['month'], x['year']), column]
col_name = "norm" + str(column)
x[col_name] = x[column] / mean_col # norm
return x
df2 = df.apply(find_norm, df_col_list = ['A','B','C'], axis=1)
#df2 will now have 3 additional columns - normA, normB, normC
df2:
A B C grp month year normA normB normC
2019-02-01 09:30:07 1 2 3 1 2 2019 0.666667 0.8 1.5
2019-03-02 09:30:07 2 3 4 1 3 2019 1.000000 1.0 1.0
2019-02-01 09:40:07 2 3 1 2 2 2019 1.000000 1.0 1.0
2019-02-01 09:38:07 2 3 1 1 2 2019 1.333333 1.2 0.5
Alternatively, for step 3, one can join the agg and df dataframes and find the norm.
Hope this helps!
Here is how the code would look like:
# Step 1
df['month'] = df.index.month
df['year'] = df.index.year # added year assuming the grouping occurs
# Step 2
agg = df.groupby(by=['grp', 'month', 'year'], as_index=True).mean()
# Step 3
def find_norm(x, df_col_list): # x is a row in dataframe, col_list is the list of columns to normalize
for column in df_col_list: # iterate over col list, find mean from aggregations, and divide the value by the mean.
mean_col = agg.loc[(str(x['grp']), x['month'], x['year']), column]
col_name = "norm" + str(column)
x[col_name] = x[column] / mean_col # norm
return x
df2 = df.apply(find_norm, df_col_list = ['A','B','C'], axis=1)
I'm trying to modify multiple column values in pandas.Dataframes with different increments in each column so that the values in each column do not overlap with each other when graphed on a line graph.
Here's the end goal of what I want to do: link
Let's say I have this kind of Dataframe:
Col1 Col2 Col3
0 0.3 0.2
1 1.1 1.2
2 2.2 2.4
3 3 3.1
but with hundreds of columns and thousands of values.
When graphing this on a line-graph on excel or matplotlib, the values overlap with each other, so I would like to separate each column by adding the same values for each column like so:
Col1(+0) Col2(+10) Col3(+20)
0 10.3 20.2
1 11.1 21.2
2 12.2 22.4
3 13 23.1
By adding the same value to one column and increasing by an increment of 10 over each column, I am able to see each line without it overlapping in one graph.
I thought of using loops and iterations to automate this value-adding process, but I couldn't find any previous solutions on Stackoverflow that addresses how I could change the increment value (e.g. from adding 0 in Col1 in one loop, then adding 10 to Col2 in the next loop) between different columns, but not within the values in a column. To make things worse, I'm a beginner with no clue about programming or data manipulation.
Since the data is in a CSV format, I first used Pandas to read it and store in a Dataframe, and selected the columns that I wanted to edit:
import pandas as pd
#import CSV file
df = pd.read_csv ('data.csv')
#store csv data into dataframe
df1 = pd.DataFrame (data = df)
# Locate columns that I want to edit with df.loc
columns = df1.loc[:, ' C000':]
here is where I'm stuck:
# use iteration with increments to add numbers
n = 0
for values in columns:
values = n + 0
print (values)
But this for-loop only adds one increment value (in this case 0), and adds it to all columns, not just the first column. Not only that, but I don't know how to add the next increment value for the next column.
Any possible solutions would be greatly appreciated.
IIUC ,just use df.add() over axis=1 with a list made from the length of df.columns:
df1 = df.add(list(range(0,len(df.columns)*10))[::10],axis=1)
Or as #jezrael suggested, better:
df1=df.add(range(0,len(df.columns)*10, 10),axis=1)
print(df1)
Col1 Col2 Col3
0 0 10.3 20.2
1 1 11.1 21.2
2 2 12.2 22.4
3 3 13.0 23.1
Details :
list(range(0,len(df.columns)*10))[::10]
#[0, 10, 20]
I would recommend you to avoid looping over the data frame as it is inefficient but rather think of adding to matrixes.
e.g.
import numpy as np
import pandas as pd
# Create your example df
df = pd.DataFrame(data=np.random.randn(10,3))
# Create a Matrix of ones
x = np.ones(df.shape)
# Multiply each column with an incremented value * 10
x = x * 10*np.arange(1,df.shape[1]+1)
# Add the matrix to the data
df + x
Edit: In case you do not want to increment with 10, 20 ,30 but 0,10,20 use this instead
import numpy as np
import pandas as pd
# Create your example df
df = pd.DataFrame(data=np.random.randn(10,3))
# Create a Matrix of ones
x = np.ones(df.shape)
# THIS LINE CHANGED
# Obmit the 1 so there is only an end value -> default start is 0
# Adjust the length of the vector
x = x * 10*np.arange(df.shape[1])
# Add the matrix to the data
df + x