Rolling average of pandas timeseries with fixed duration window - pandas

In the docs it says
df.rolling('2s').sum()
is an option. However
ts = ts.rolling('900s').mean()
returns the error ValueError: window must be an integer
What's up with that? Can't share the whole timeseries, but this is a snippet:
2019-09-20 00:30:51.234260+00:00 4.63
2019-09-20 00:50:51.292892+00:00 4.40
2019-09-20 01:30:51.273058+00:00 4.54
2019-09-20 01:50:51.270876+00:00 4.44
2019-09-20 02:30:51.267385+00:00 4.55
...
2019-10-30 22:57:35.003066+00:00 1.71
2019-10-30 23:12:34.965801+00:00 1.61
2019-10-30 23:27:34.976495+00:00 1.56
2019-10-30 23:42:34.984976+00:00 1.26
2019-10-30 23:57:34.965543+00:00 1.05

Related

pandas (multi) index wrong need to change it

I have a DataFrame multiData that looks like this:
print(multiData)
Date Open High Low Close Adj Close Volume
Ticker Date
AAPL 0 2010-01-04 7.62 7.66 7.59 7.64 6.51 493729600
1 2010-01-05 7.66 7.70 7.62 7.66 6.52 601904800
2 2010-01-06 7.66 7.69 7.53 7.53 6.41 552160000
3 2010-01-07 7.56 7.57 7.47 7.52 6.40 477131200
4 2010-01-08 7.51 7.57 7.47 7.57 6.44 447610800
... ... ... ... ... ... ... ...
META 2668 2022-12-23 116.03 118.18 115.54 118.04 118.04 17796600
2669 2022-12-27 117.93 118.60 116.05 116.88 116.88 21392300
2670 2022-12-28 116.25 118.15 115.51 115.62 115.62 19612500
2671 2022-12-29 116.40 121.03 115.77 120.26 120.26 22366200
2672 2022-12-30 118.16 120.42 117.74 120.34 120.34 19492100
I need to get rid of "Date 0, 1, 2, ..." column and make the actual "Date" column part of the (multi) index
How do I do this?
Use df.droplevel to delete level 1 and chain df.set_index to add column Date to the index by setting the append parameter to True.
df = df.droplevel(1).set_index('Date', append=True)
df
Open High Low Close Adj Close Volume
Ticker Date
AAPL 2010-01-04 7.62 7.66 7.59 7.64 6.51 493729600
2010-01-05 7.66 7.70 7.62 7.66 6.52 601904800

missing data in pandas web.DataReader yahoo

I am collecting data from yahoo, it works fine for more than 3 months. Then in last week I noticed that datas from 22.4.2021 is missing till today.
When I print the stockData, here is the last records :
...
...
2021-04-20 7.52 7.07 7.35 7.13 4119197.0 7.13
2021-04-21 7.36 7.13 7.16 7.25 3110870.0 7.25
2021-04-22 7.73 7.10 7.22 7.59 13178439.0 7.59
2021-05-06 10.08 9.48 9.52 9.93 2753885.0 9.93
As you see, datas are missing. It's because of yahoo or panda function, Could you help me to fix it ?.
Alper

Resample a datetimeIndex start day wrong

Source:
import pandas as pd
import numpy as np
cols = ['Date', 'Time', 'Load', 'Battery', 'Panel',
'Wind', 'Temp', 'Humidity', 'Volt']
data = pd.read_csv('test.csv',delimiter=';',header=0,names=cols,
decimal=',',parse_dates[[0,1]],
infer_datetime_format=True)
data.set_index('Date_Time',inplace=True)
I have this data frame:
In [126]: data.head()
Out[126]:
Load Battery Panel Wind Temp Humidity Volt
Date_Time
2018-07-31 13:07:15 13.3 326.3 353.1 0.98 33.93 21.92 3.89
2018-07-31 13:08:15 14.0 314.4 342.5 0.59 33.88 21.84 3.88
2018-07-31 13:09:16 13.4 309.6 335.5 0.39 33.84 22.14 3.88
2018-07-31 13:10:16 13.8 285.1 313.8 2.55 33.71 23.18 3.88
2018-07-31 13:11:16 13.6 292.9 314.7 2.03 33.62 23.25 3.88
......
with other 93000 rows. from 2018-07-31 to 2018-04-10. I'd like to resample by taking the sum of values for each 10minute frame. So I tried:
In [127]: data.resample('10min',closed='left',label='left').sum()
Out[127]:
Load Battery Panel Wind Temp Humidity Volt
Date_Time
2018-01-08 00:00:00 136.9 -140.6 -2.9 19.06 291.27 245.63 39.45
2018-01-08 00:10:00 137.3 -140.7 -3.1 15.14 290.62 244.88 39.42
2018-01-08 00:20:00 137.4 -140.4 -2.3 18.03 288.61 246.44 39.44
2018-01-08 00:30:00 137.5 -140.4 -2.2 12.61 286.97 246.83 39.43
That is close to what I expect, but the 'resample' remove all the data from the first day (I suspect maybe because the series do not start at midnight), what is the proper way to do the resampling? There are two issues:
The first day is missing in the result, i.e. all data removed and the resampled dataframe starts in the first of august and not on 07/31.
It is ok to consider intervals that starts at midnight and are so, perfectly multiple of 10min (so, ok for 00:00, 10:00, 20:00) but then I expect that the first grouping is:
2018-07-31 13:07:15 13.3 326.3 353.1 0.98 33.93 21.92 3.89
2018-07-31 13:08:15 14.0 314.4 342.5 0.59 33.88 21.84 3.88
2018-07-31 13:09:16 13.4 309.6 335.5 0.39 33.84 22.14 3.88
and then from 13:10:16, of course in the first day of the dataset and not on the second.
Ok. I solved it using:
x = data['2018-07-31'].resample('10min').sum()
y = data.resample('10min',closed='left',label='left').sum()
r = pd.concat([x,y])
but I think that this must be a form of bug in resample.
For output that starts at exactly 2018-07-31 13:07:15, you need to add in the argument base: "the origin of the aggregated intervals": documentation.
Example code:
start = pd.to_datetime('2018-07-31 13:07:15', format='%Y-%m-%d %H:%M:%S')
minutes = pd.date_range(start, start + timedelta(10), freq='min')
df = pd.DataFrame({'Date_Time': minutes, 'Load': np.random.randint(13, size=len(minutes))})
df.set_index('Date_Time', inplace=True)
df.resample('10min', closed='left', label='left', base=7.25).sum()
Result:
Date_Time Load
2018-07-31 13:07:15 11
2018-07-31 13:17:15 1
2018-07-31 13:27:15 6

Data analysis over multiple dataframes? Panels or Multi-index?

When I extract data for multiple stocks using web.DataReader, I'm getting a panel as the output.
import numpy as np
import pandas as pd
import pandas_datareader.data as web
import matplotlib.pyplot as plt
import datetime as dt
import re
startDate = '2010-01-01'
endDate = '2016-09-07'
stocks_query = ['AAPL','OPK']
stocks = web.DataReader(stocks_query, data_source='yahoo',
start=startDate, end=endDate)
stocks = stocks.swapaxes('items','minor_axis')
Leading to an output of :
Dimensions: 2 (items) x 1682 (major_axis) x 6 (minor_axis)
Items axis: AAPL to OPK
Major_axis axis: 2010-01-04 00:00:00 to 2016-09-07 00:00:00
Minor_axis axis: Open to Adj Close
A single dataframe of the panel looks like this
stocks['OPK']
Open High Low Close Volume Adj Close log_return \
Date
2010-01-04 1.80 1.97 1.76 1.95 234500.0 1.95 NaN
2010-01-05 1.64 1.95 1.64 1.93 135800.0 1.93 -0.010309
2010-01-06 1.90 1.92 1.77 1.79 546600.0 1.79 -0.075304
2010-01-07 1.79 1.94 1.76 1.92 138700.0 1.92 0.070110
2010-01-08 1.92 1.94 1.86 1.89 62500.0 1.89 -0.015748
I plan to do a lot of data manipulation across all dataframes adding new columns. comparing two dataframes etc. I was recommended to looking into multi_indexing, as panels are being deprecated.
This is my fist time working with panels.
If I want to add a new column to both dataframes (AAPL, OPK), I had to do something like this:
for i in stocks:
stocks[i]['log_return'] = np.log(stocks[i]['Close']/(stocks[i]['Close'].shift(1)))
If multi_indexing is indeed recommended for working with multiple dataframes, how exactly would I convert my dataframes into a form I can easily work with?
Would I have one main index, with the next-level being the stocks, and the columns would be contained within each stock?
I went through the docs, which gave many examples using tuples o which I didn't get or examples using single dataframes.
http://pandas.pydata.org/pandas-docs/stable/advanced.html
So how exactly do I convert my panel into a multi_index dataframe?
I'd like to extend #piRSquared's answer with some examples:
In [40]: stocks.to_frame()
Out[40]:
AAPL OPK
Date minor
2010-01-04 Open 2.134300e+02 1.80
High 2.145000e+02 1.97
Low 2.123800e+02 1.76
Close 2.140100e+02 1.95
Volume 1.234324e+08 234500.00
Adj Close 2.772704e+01 1.95
2010-01-05 Open 2.146000e+02 1.64
High 2.155900e+02 1.95
Low 2.132500e+02 1.64
Close 2.143800e+02 1.93
... ... ...
2016-09-06 Low 1.075100e+02 9.19
Close 1.077000e+02 9.36
Volume 2.688040e+07 3026900.00
Adj Close 1.066873e+02 9.36
2016-09-07 Open 1.078300e+02 9.39
High 1.087600e+02 9.60
Low 1.070700e+02 9.38
Close 1.083600e+02 9.59
Volume 4.236430e+07 2632400.00
Adj Close 1.073411e+02 9.59
[10092 rows x 2 columns]
but if you want to convert it to MultiIndex DF, it's better to leave the original pandas_datareader Panel as it is:
In [38]: p = web.DataReader(stocks_query, data_source='yahoo', start=startDate, end=endDate)
In [39]: p.to_frame()
Out[39]:
Open High Low Close Volume Adj Close
Date minor
2010-01-04 AAPL 213.429998 214.499996 212.380001 214.009998 123432400.0 27.727039
OPK 1.800000 1.970000 1.760000 1.950000 234500.0 1.950000
2010-01-05 AAPL 214.599998 215.589994 213.249994 214.379993 150476200.0 27.774976
OPK 1.640000 1.950000 1.640000 1.930000 135800.0 1.930000
2010-01-06 AAPL 214.379993 215.230000 210.750004 210.969995 138040000.0 27.333178
OPK 1.900000 1.920000 1.770000 1.790000 546600.0 1.790000
2010-01-07 AAPL 211.750000 212.000006 209.050005 210.580000 119282800.0 27.282650
OPK 1.790000 1.940000 1.760000 1.920000 138700.0 1.920000
2010-01-08 AAPL 210.299994 212.000006 209.060005 211.980005 111902700.0 27.464034
OPK 1.920000 1.940000 1.860000 1.890000 62500.0 1.890000
... ... ... ... ... ... ...
2016-08-31 AAPL 105.660004 106.570000 105.639999 106.099998 29662400.0 105.102360
OPK 9.260000 9.260000 9.070000 9.100000 2793300.0 9.100000
2016-09-01 AAPL 106.139999 106.800003 105.620003 106.730003 26701500.0 105.726441
OPK 9.310000 9.540000 9.190000 9.290000 3515300.0 9.290000
2016-09-02 AAPL 107.699997 108.000000 106.820000 107.730003 26802500.0 106.717038
OPK 9.340000 9.390000 9.160000 9.330000 2061200.0 9.330000
2016-09-06 AAPL 107.900002 108.300003 107.510002 107.699997 26880400.0 106.687314
OPK 9.320000 9.480000 9.190000 9.360000 3026900.0 9.360000
2016-09-07 AAPL 107.830002 108.760002 107.070000 108.360001 42364300.0 107.341112
OPK 9.390000 9.600000 9.380000 9.590000 2632400.0 9.590000
[3364 rows x 6 columns]
How to work with a MultiIndex DF:
In [46]: df = p.to_frame()
In [47]: df.loc[pd.IndexSlice[:, ['AAPL']], :]
Out[47]:
Open High Low Close Volume Adj Close
Date minor
2010-01-04 AAPL 213.429998 214.499996 212.380001 214.009998 123432400.0 27.727039
2010-01-05 AAPL 214.599998 215.589994 213.249994 214.379993 150476200.0 27.774976
2010-01-06 AAPL 214.379993 215.230000 210.750004 210.969995 138040000.0 27.333178
2010-01-07 AAPL 211.750000 212.000006 209.050005 210.580000 119282800.0 27.282650
2010-01-08 AAPL 210.299994 212.000006 209.060005 211.980005 111902700.0 27.464034
2010-01-11 AAPL 212.799997 213.000002 208.450005 210.110003 115557400.0 27.221758
2010-01-12 AAPL 209.189995 209.769995 206.419998 207.720001 148614900.0 26.912110
2010-01-13 AAPL 207.870005 210.929995 204.099998 210.650002 151473000.0 27.291720
2010-01-14 AAPL 210.110003 210.459997 209.020004 209.430000 108223500.0 27.133657
2010-01-15 AAPL 210.929995 211.599997 205.869999 205.930000 148516900.0 26.680198
... ... ... ... ... ... ...
2016-08-24 AAPL 108.570000 108.750000 107.680000 108.029999 23675100.0 107.014213
2016-08-25 AAPL 107.389999 107.879997 106.680000 107.570000 25086200.0 106.558539
2016-08-26 AAPL 107.410004 107.949997 106.309998 106.940002 27766300.0 105.934466
2016-08-29 AAPL 106.620003 107.440002 106.290001 106.820000 24970300.0 105.815591
2016-08-30 AAPL 105.800003 106.500000 105.500000 106.000000 24863900.0 105.003302
2016-08-31 AAPL 105.660004 106.570000 105.639999 106.099998 29662400.0 105.102360
2016-09-01 AAPL 106.139999 106.800003 105.620003 106.730003 26701500.0 105.726441
2016-09-02 AAPL 107.699997 108.000000 106.820000 107.730003 26802500.0 106.717038
2016-09-06 AAPL 107.900002 108.300003 107.510002 107.699997 26880400.0 106.687314
2016-09-07 AAPL 107.830002 108.760002 107.070000 108.360001 42364300.0 107.341112
[1682 rows x 6 columns]
You're going to love this one
stocks.to_frame()

seasonal_decompose: operands could not be broadcast together with shapes on a series

I know there are many questions on this topic, but none of them helped me to solve this problem. I'm really stuck on this.
With a simple series:
0
2016-01-31 266
2016-02-29 235
2016-03-31 347
2016-04-30 514
2016-05-31 374
2016-06-30 250
2016-07-31 441
2016-08-31 422
2016-09-30 323
2016-10-31 168
2016-11-30 496
2016-12-31 303
import statsmodels.api as sm
logdf = np.log(df[0])
decompose = sm.tsa.seasonal_decompose(logdf,freq=12, model='additive')
decomplot = decompose.plot()
i keep getting: ValueError: operands could not be broadcast together with shapes (12,) (14,)
I've tried pretty much everything, passing only logdf.values, passing a non-log series. It doesn't work.
Numpy and statsmodel versions:
print(statsmodels.__version__)
print(pd.__version__)
print(np.__version__)
0.6.1
0.18.1
1.11.3
As #yoonforh pointed, in my case this was fixed by setting the freq parameter to less than the time series length. E.g. if your time series ts looks like this:
2014-01-01 0.0
2014-02-01 0.0
2014-03-01 1.0
2014-04-01 1.0
2014-05-01 0.0
2014-06-01 1.0
2014-07-01 1.0
2014-08-01 0.0
2014-09-01 0.0
2014-10-01 1.0
2014-11-01 0.0
2014-12-01 0.0
the shape is
(12,)
so this will give the error as per above:
seasonal_decompose(ts, freq=12, model='additive')
but if I try freq=11 or any other int less than 12, e.g.
seasonal_decompose(ts, freq=11, model='additive')
this works
i noticed that with newer pandas and statsmodel versions it seems to work.
Given a series:
2016-01-03 8.326275
2016-01-10 8.898229
2016-01-17 8.754792
2016-01-24 8.658172
2016-01-31 8.731659
2016-02-07 9.047233
2016-02-14 8.799662
2016-02-21 8.783549
2016-02-28 8.782783
2016-03-06 9.081825
2016-03-13 8.737934
2016-03-20 8.658693
2016-03-27 8.666475
2016-04-03 9.029178
2016-04-10 8.781555
2016-04-17 8.720787
2016-04-24 8.633909
2016-05-01 8.937744
2016-05-08 8.804925
2016-05-15 8.766862
2016-05-22 8.651899
2016-05-29 8.653645
...
And pd/sm version:
statsmodels.__version__ 0.8.0
pandas.__version__ 0.20.1
This is the result:
import statsmodels.api as sm
logdf = np.log(df_series)
decompose = sm.tsa.seasonal_decompose(logdf, model='additive', filt=None, freq=1, two_sided=True)
decompose.plot()
I hope this could solve your problem too.