facebookresearch / kats Goto Github PK

Why does Kats has a fix dependency to fbprophet 0.7 ?

It downgraded our version from 0.7.0 to to 0.7.1 and therefore reverted our backend to pystan instead of cdmstanpy.

Hi, related to other issues mentioned (downlgrade of fbprohet to 0.7 for example) why is Kats not using the version 1.0.1 of prophet (new name of fbprophet) ?

I have tried running MKDectector on monthly values, but the algorithm is obviously limited to daily data although this limitation is not mentioned (for instance, CUSUM works on monthly data).

What a pity not to be able to process the monthly data. I would like you to add this functionality, in particular concerning the treatment of hydroclimatic data for the challenges of climate change.

Thank you for sharing this great tool

Hi FB team,

Thank you for open-sourcing your great package.
I would like to know what is the difference between Kats and Sktime:

Sktime - A unified framework for machine learning with time series

Which features Kats offers, which are not included in Sktime?

Thank you,
Eli

Kats website sharing Google analytics ID with hydra.cc (UA-149862507-1).
From the source of https://facebookresearch.github.io/Kats/:

Please use your own Google analytics ID, you are messing up the analytics of hydra.cc.

Thanks.

Hi there!

I was wondering to contribute and fix a minor style problem in homepage.

The index has a main section tag with a "features_3azU" class. This section has 2 rows:

• the first one is a generic description of the package;
• the second one contains all the 4 features divided by a col col--3 class.

For the styling I think could be better to put the description paragraph inside a col col--12 class.
In this way this section will have a 16px right and left padding and will be better aligned to the features in the following row.

I saw the CI failing because of a library called deprecated which is missing from the requirements.txt

Made a PR that adds the requirement.

Should these requirements be separate from the requirements.txt and create a test_requirements.txt or all of the dependencies for testing and development are stated at requirements.txt?

The CUSUMDetector and the RobustStatDetector both reference the name of the TimeSeriesData value column and have errors when the series name is not set.

The series name can be not set through initializing TimeSeriesData without using a DataFrame

import pandas as pd
import numpy as np
from kats.consts import TimeSeriesData

n = 10
time = pd.Series(pd.date_range(start='2018-01-01', periods=n, freq='D'))
value = pd.Series(np.arange(n))
ts = TimeSeriesData(time=time, value=value)

print("ts.value.name", ts.value.name) # prints None
print("ts.time.name", ts.time.name) # prints None

This causes both to error out when attempting to plot and the RobustStatDetector to error out when detecting change points.

For calling the .plot() method on either the error is:

Traceback (most recent call last):
  File "/home/jrand/workspace/kats/.venv/lib/python3.8/site-packages/pandas/core/indexes/base.py", line 3361, in get_loc
    return self._engine.get_loc(casted_key)
  File "pandas/_libs/index.pyx", line 76, in pandas._libs.index.IndexEngine.get_loc
  File "pandas/_libs/index.pyx", line 108, in pandas._libs.index.IndexEngine.get_loc
  File "pandas/_libs/hashtable_class_helper.pxi", line 5198, in pandas._libs.hashtable.PyObjectHashTable.get_item
  File "pandas/_libs/hashtable_class_helper.pxi", line 5206, in pandas._libs.hashtable.PyObjectHashTable.get_item
KeyError: None

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "error.py", line 19, in <module>
    detector.plot(change_points)
  File "/home/jrand/workspace/kats/kats/detectors/cusum_detection.py", line 565, in plot
    plt.plot(data_df[time_col_name], data_df[val_col_name])
  File "/home/jrand/workspace/kats/.venv/lib/python3.8/site-packages/pandas/core/frame.py", line 3455, in __getitem__
    indexer = self.columns.get_loc(key)
  File "/home/jrand/workspace/kats/.venv/lib/python3.8/site-packages/pandas/core/indexes/base.py", line 3363, in get_loc
    raise KeyError(key) from err
KeyError: None

When calling the RobustStatDetector.detector() the error is:

Traceback (most recent call last):
  File "error.py", line 22, in <module>
    change_points = detector.detector()
  File "/home/jrand/workspace/kats/kats/detectors/robust_stat_detection.py", line 58, in detector
    data_df = data_df.set_index(time_col_name)
  File "/home/jrand/workspace/kats/.venv/lib/python3.8/site-packages/pandas/util/_decorators.py", line 311, in wrapper
    return func(*args, **kwargs)
  File "/home/jrand/workspace/kats/.venv/lib/python3.8/site-packages/pandas/core/frame.py", line 5446, in set_index
    raise KeyError(f"None of {missing} are in the columns")
KeyError: 'None of [None] are in the columns'

Both of these issues can be fixed fairly easily.

Checking if the values are named and setting a default if they are not. This can be done by adding the following code to each of their __init__ functions.

if self.data.value.name is None:
    self.data.value.name = DEFAULT_VALUE_NAME

if self.data.time.name is None:
    self.data.time.name = DEFAULT_TIME_NAME

Or they can be refactored to not use the names at all. As both are univariate only detectors we can refer to the columns simply as self.data.time and self.data.value without the need for using the column name specifically.

1. My environment is Win 10, with conda 4.9.2 and Python 3.7.6
2. While trying to install Kats (pip install kats), I am getting the error: Building wheel for fbprophet (setup.py) ... error
3. More lines from the error message are in attached file
   Kats_error.txt
4. I had an older working fbprophet, faced same errors in trying to upgrade (pip install --upgrade fbprophet).
5. Anybody know how to fix this?

the TsFeatures.transform() seems to produce some nan values when using all feature groups , for the Holt-Winters features and the changepoint detection features

An example of the error can be found at https://colab.research.google.com/drive/13t3Gy9vazkeb8J32Zv-dijuww03GDbE7?usp=sharing, using the air passengers dataset from https://github.com/facebookresearch/Kats/blob/master/kats/data/air_passengers.csv, and stock price data from the yfinance library

Hi there,
nice library, would be nice to have a module for finding spikes/peaks like this one: https://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.find_peaks.html

Cheers.

Hi, kudos to your work as this package will be quite useful. However, it was quite a bad surprise for me to add this package to my project dependencies and see huge and non-environment-friendly packages such as pytorch, gpytorch, llvmlite and even plotly being installed into my environment. I only plan to use some changepoint detection methods and maybe visualizations, and I highly doubt that (though, of course correct me if I am wrong) the whole torch stack is necessary for such a task. Similarly, I imagine plotly can easily be made an optional dependency as well, since matplotlib is still the de facto plotting library of python and it would not be in the interest of majority of the users to install a data analysis package and get a big python-js bundle with it.

I have time series data with the time_col is the index of the dataframe:

df = [index  value 
            0         31
            1          22
            2         15
            3         77]

When I am trying to convert it to TimeSeriesData, the index is automatically transform to epoch time. (e.g. 1970-01-01 00:00:00.000000095)
Is there a way to keep the time_col as the mere index when using TimeSeriesData

Hello,
First of all thanks for this amazing work! It's the lib I was looking for for so long.
Manipulating multivariate time series can still be quite frustrating.
Have you plan to add getitem method to TimeSeriesData object to enable pandas-like column access df["column_name"] ?
It could be especially convenient to get a univariate TimeSeriesData object ready to be used in Kats models.

Meanwhile any workaround more convinient than the following would be welcome:

uni_ts = TimeSeriesData(multi_ts.value["col_name"])

Best regards,

I have a pandas dataframe dat with two columns, the first one having timestamps, named 'time', and the second one containing the data values, named 'value'. The timestamps have a frequency of 5 min i.e. the values there are of the type "2018-01-01 00:00:00", "2018-01-01 00:05:00", "2018-01-01 00:10:00", etc.

I invoked kats.TimeSeriesData() on this dataset, then apply outlier detection.
Code run by me:
dat['time'] = pd.to_datetime(dat['time']) #converting timestamp column to pandas datetime
dat_ts = TimeSeriesData(dat)
from kats.detectors.outlier import OutlierDetector
ts_outlierDetection = OutlierDetector(dat_ts, 'additive') # call OutlierDetector
ts_outlierDetection.detector()

I get the following error on running the last line.

ERROR:root:!! Traceback (most recent call last):
!! File ".../kats/detectors/outlier.py", line 114, in detector
outlier = self.__clean_ts__(ts)
!! File ".../kats/detectors/outlier.py", line 88, in __clean_ts__
result = seasonal_decompose(original, model=self.decomp)
!! File ".../pandas/util/_decorators.py", line 207, in wrapper
return func(*args, **kwargs)
!! File ".../statsmodels/tsa/seasonal.py", line 140, in seasonal_decompose
pfreq = freq_to_period(pfreq)
!! File ".../statsmodels/tsa/tsatools.py", line 829, in freq_to_period
raise ValueError("freq {} not understood. Please report if you "
!! ValueError: freq T not understood. Please report if you think this is in error.
ERROR:root:Outlier Detection Failed

I looked into line 829 of tsatools.py and it seems it can only support few timestamp formats and hence not recognizing the 5 min frequency in my timeseries data? Should I be setting something like freq='5min' somewhere in the codes?

I also ran the outlier detection code on daily data, i.e. freq = 'D', and there seems to be no issues (this is essentially the example in the tutorial).

Hello!

When I was trying to import some kats functionalities on my Jupyter Notebook, it came back with a error message as below:

ImportError Traceback (most recent call last)
in
----> 1 from kats.detectors.outlier import OutlierDetector

~\Anaconda3\lib\site-packages\kats_init_.py in
1 from . import consts # noqa
----> 2 from . import utils # noqa
3 from . import detectors # noqa
4 from . import models # noqa
5 from . import tsfeatures # noqa

ImportError: cannot import name 'utils' from partially initialized module 'kats' (most likely due to a circular import) (C:\Users\49683\Anaconda3\lib\site-packages\kats_init_.py)

Anyone knows how to resolve this issue? Thanks in advance!

Hi, first, Thanks for your awesome time-series analysis framework for everyone!
I am following your tutorial of Kats(101_basics) but I wondered what meaning of argument freq="MS" in predict method. What is the meaning of freq and how can I set the other freq parameter in addtition to "MS"?

Below, here is the screenshot of your tutorial, Thanks!

Thank you for putting together this really awesome time series analysis package. Given that the aim of Kats is to provide a one-stop-shop for time series analysis, I was wondering if there are any plans/interest to incorporate time series analyses with matrix profiles using the STUMPY package. Given that STUMPY only depends on NumPy, SciPy, and Numba (all of which are Kats dependencies) and it is has 100% code coverage, it may be a great complement to the Kats ecosystem.

Full disclosure, I am the creator and core maintainer of the STUMPY package. Thank you for your consideration!

Hi. thanks for your awesome project!
I am following your tutorial and learning Kats library for multi-variate forecasting. But in the tutorial(201_forecasting.ipynb), I learned multi-variate forecasting(VAR model in tutorial) and I have a question about this. As usual, multi-variate forecasting means to predict only one target value(Y variable) using multi features(X variables). But this tutorial doesn't provide me with the multi-variate forecasting that I want. This tutorial(VAR model) say me that this is multi-variate forecasting but it has two X features and finally predicts two Y target values. How can I predict multi-variate forecasting that I want? Or VAR model can't be used to predict only one Y target value using more than two X features?

from kats.models.theta import ThetaModel, ThetaParams
params = ThetaParams(m=7) # Weekly seasonality 
m = ThetaModel(data_ts, params) # data_ts -> daily data points
m.fit()

gives this error:

Optimization failed to converge. Check mle_retvals.

---------------------------------------------------------------------------
LinAlgError                               Traceback (most recent call last)
/var/folders/25/t45ty5ps70l5y9gww_k11zy40000gq/T/ipykernel_44596/687703454.py in <module>
----> 1 m.fit()

~/kats/lib/python3.9/site-packages/kats/models/theta.py in fit(self, **kwargs)
    119         # creating x and intercept variables to fit a straight line
    120         regr = np.vstack([np.arange(self.n), np.ones(self.n)]).T
--> 121         slope, _ = np.linalg.lstsq(regr, deseas_data.value.values)[0]
    122         # pyre-fixme[16]: `ThetaModel` has no attribute `drift`.
    123         self.drift = slope / 2

<__array_function__ internals> in lstsq(*args, **kwargs)

~/kats/lib/python3.9/site-packages/numpy/linalg/linalg.py in lstsq(a, b, rcond)
   2273     m2, n_rhs = b.shape[-2:]
   2274     if m != m2:
-> 2275         raise LinAlgError('Incompatible dimensions')
   2276 
   2277     t, result_t = _commonType(a, b)

LinAlgError: Incompatible dimensions

Ran the same data under Prophet model and it fits and forecasts w/o errors. Any pointers as to the source of the problem?

I get the error below when I run the tutorial kats_202_detection.ipynb https://github.com/facebookresearch/Kats/blob/master/tutorials/kats_202_detection.ipynb
Any clue ?

KeyError: Timestamp('2019-12-23 23:59:58.142906')

The above exception was the direct cause of the following exception:

KeyError                                  Traceback (most recent call last)
<ipython-input-265-b73347c84449> in <module>
      3 d = MultivariateAnomalyDetector(multi_anomaly_ts, params, training_days=60)
      4 display(params)
----> 5 anomaly_score_df = d.detector()
      6 
      7 d.plot()

~\AppData\Local\Continuum\anaconda3\lib\site-packages\kats\detectors\outlier.py in detector(self)
    300         while fcstTime < self.df.index.max():
    301             # forecast for fcstTime+ 1
--> 302             pred_df = self._generate_forecast(fcstTime)
    303             # calculate anomaly scores
    304             anomaly_scores_t = self._calc_anomaly_scores(pred_df)

How can I use exogenous variable with ProphetModel?

When running the kats_101_basics.ipynb, it has an error shown as following

from dateutil import parser
from datetime import datetime

# Convert time from air_passengers data to unix time
air_passengers_ts_unixtime = air_passengers_df.time.apply(
    lambda x: datetime.timestamp(parser.parse(x))
)

air_passengers_ts_unixtime

---------------------------------------------------------------------------
OSError                                   Traceback (most recent call last)
<ipython-input-28-38e65b10d87c> in <module>
      4 # Convert time from air_passengers data to unix time
      5 air_passengers_ts_unixtime = air_passengers_df.time.apply(
----> 6     lambda x: datetime.timestamp(parser.parse(x))
      7 )
      8 

~\Anaconda3\envs\fb\lib\site-packages\pandas\core\series.py in apply(self, func, convert_dtype, args, **kwds)
   4198             else:
   4199                 values = self.astype(object)._values
-> 4200                 mapped = lib.map_infer(values, f, convert=convert_dtype)
   4201 
   4202         if len(mapped) and isinstance(mapped[0], Series):

pandas\_libs\lib.pyx in pandas._libs.lib.map_infer()

<ipython-input-28-38e65b10d87c> in <lambda>(x)
      4 # Convert time from air_passengers data to unix time
      5 air_passengers_ts_unixtime = air_passengers_df.time.apply(
----> 6     lambda x: datetime.timestamp(parser.parse(x))
      7 )
      8 

OSError: [Errno 22] Invalid argument

Hi all,

to my knowledge and research within the kats modules it is not possible to add additional regressors to a ProphetModel as of now, correct?

I am refering to this functionality: https://facebook.github.io/prophet/docs/seasonality,_holiday_effects,_and_regressors.html#additional-regressors

Will there be a possibility to add such functionality to kats implementation of Prophet?
I am think of something similar to custom_seasonalities like here:

Adding a new optinal parameter to the constructor e.g. additional_regressors: Optional[pd.DataFrame]` with the same time and the regressors columns.

Would appreciate some feedback on this idea.

It would be nice to add a Kats package to conda-forge https://conda-forge.org (prophet is already available in conda-forge: https://github.com/conda-forge/prophet-feedstock).

The "prophet" conda package should probably be used as dependency (instead of the "fbprophet" package), also see #62

First of all, great package --- kudos to the team that created it!

I was trying to adopt the grid search code in tutorial 201 , which introduces forecasting, for my own needs. In particular, I ran into two problems --- the resolution of which may benefit from more comprehensive docs.

As an example, here is how I defined the search space for tuning the hyperparameters of VARModel.

parameters_grid_search = [
    {
        "name": "maxlags",
        "type": "choice",
        "values": list(range(1, 14)),
        "value_type": "int",
        "is_ordered": True
    },
]

First, it is unclear what are the valid options for type, and what different effects they have. Second, it is unclear that values have to be a list, not e.g., a tuple --- until I ran into an AssertionError.

The first one could really use input from people familiar with kats' tuning backend. The second perhaps can be remedied by adding a note in the tutorial notebook and/or API docs; would anyone mind me making a pull request?

Hi,

Is the TSfeatures available in Kats same as the tsfeatures package (https://pypi.org/project/tsfeatures/) ? Have you done any benchmarking against that package or the original tsfeatures in R?

For example, the link above also shows "Comparison with the R implementation (sum of absolute differences)". It would be great to have a similar comparison on M3/M4 data.

Thanks,

Reported by Xiaodong:

Something caused some of the docs to get obliterated in change 356718e
For example, tsfeature.tsfeatures

"All models are not working. Detection works well"

The way some of the plotting is done right now is very limiting.
For instance kats/detectors/cusum_detection.py ends with a plt.show().

It would be better to allow the caller to manage the figure itself and pass an axis (as in most typical graphing utilities). This way one could build composite figures or decide where to plot (to an html file for instance).

Would you consider publishing a wheel alongside the source distribution? I understand Kats is Python source-only, but it would help to publish a *-py3-none-any.whl file using bdist_wheel.

Hi all,

thanks for a great package. Sorry to ask a question here (not necessarily and issue yet) but can't find the users group

I have a quick question, is it possible to update a Prophet model as new data arrives or is retraining from scratch the only alternative?

Say you have a df like this

from kats.models.prophet import ProphetModel, ProphetParams
from kats.consts import TimeSeriesData
import pandas as pd
import numpy as np

df = pd.DataFrame({"time": pd.date_range('2020-01-01', '2020-01-21'), "value": np.random.randn(21)})
ts = TimeSeriesData(df)

model = ProphetModel(ts, ProphetParams())
model.fit()

and latter on we add another datapoint to df as in

df = df.append({"time": pd.to_datetime('2020-01-22'), "value": 0}, ignore_index=True)

I'm assuming that it would be possible and efficient to just update the model

Thanks for reading

Hi,
I am using MAC and python3.6.
Installed Kats using pip and when trying:
import kats

I get the error:

Traceback (most recent call last):
  File "/venv/lib/python3.6/site-packages/IPython/core/interactiveshell.py", line 3343, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-4-f96c9197f95c>", line 1, in <module>
    import kats
  File "/Applications/PyCharm.app/Contents/plugins/python/helpers/pydev/_pydev_bundle/pydev_import_hook.py", line 21, in do_import
    module = self._system_import(name, *args, **kwargs)
  File "venv/lib/python3.6/site-packages/kats/__init__.py", line 1, in <module>
    from . import consts   # noqa
  File "/Applications/PyCharm.app/Contents/plugins/python/helpers/pydev/_pydev_bundle/pydev_import_hook.py", line 21, in do_import
    module = self._system_import(name, *args, **kwargs)
  File "venv/lib/python3.6/site-packages/kats/consts.py", line 20
    from __future__ import annotations
                                     ^
SyntaxError: future feature annotations is not defined

Any idea?

Using the keyword exog for kats' SARIMA module throws an error.

Is the kats SARIMA suited for modelling exogenity?

I am pasting a snip:

If there is a way to pass exog regressors, it will be really helpful to show an example snippet.

May I suggest adding Kalman filter? Seems like it would be a good fit!

In order to force predictions to be positive (this is a common requirement for many business-related time series), the prophet package can be used with the ProphetPos class (see "Approach 5" here: facebook/prophet#1668 (comment)).

It would be nice to add a "ProphetPosModel" to Kats, so we can force predictions to be positive when we use prophet in Kats.

Maybe a positive constraint could also be added to the other models supported by Kats (SARIMA etc.).

Hello, as for prophet it would more than nice to have the ability to avoid installing the pystan software.

e.g. for fbprophet we can proceed as follows:
CMDSTAN=/tmp/cmdstan-2.22.1 STAN_BACKEND=CMDSTANPY pip install prophet

Would that be possible to install Kats the same way ? And to document in the readme ?

Hi,

I want to know that does Kats supports Multivariate time series forecasting? Please let me know

Thank you very much :)

Hi everyone,

When I try to install kats on my machine (Windows 10, build see attached screenshot), the installation breaks building the wheel for fbprophet.

I am trying this on a fresh conda environment (conda version 4.10.1) and python3.8. My only guess is that this has todo with the ability to install pystan on windows, which is not possible without access to WSL2 (which I do not have because of corporate restrictions).

I will attach the error message below.

Attachments

Windows version

Error message

ERROR: Command errored out with exit status 1:
   command: 'C:\Users\matheiss\Miniconda3\envs\tiresias-kats\python.exe' -u -c 'import io, os, sys, setuptools, tokenize; sys.argv[0] = '"'"'C:\\Users\\matheiss\\AppData\\Local\\Temp\\1\\pip-install-5dy9pwq7\\fbprophet_6fb4ff087aaf40478b71089d7e634e82\\setup.py'"'"'; __file__='"'"'C:\\Users\\matheiss\\AppData\\Local\\Temp\\1\\pip-install-5dy9pwq7\\fbprophet_6fb4ff087aaf40478b71089d7e634e82\\setup.py'"'"';f = getattr(tokenize, '"'"'open'"'"', open)(__file__) if os.path.exists(__file__) else io.StringIO('"'"'from setuptools import setup; setup()'"'"');code = f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, __file__, '"'"'exec'"'"'))' bdist_wheel -d 'C:\Users\matheiss\AppData\Local\Temp\1\pip-wheel-xzjkrfhe'
       cwd: C:\Users\matheiss\AppData\Local\Temp\1\pip-install-5dy9pwq7\fbprophet_6fb4ff087aaf40478b71089d7e634e82\
  Complete output (42 lines):
  running bdist_wheel
  running build
  running build_py
  creating build
  creating build\lib
  creating build\lib\fbprophet
  creating build\lib\fbprophet\stan_model
  Traceback (most recent call last):
    File "<string>", line 1, in <module>
    File "C:\Users\matheiss\AppData\Local\Temp\1\pip-install-5dy9pwq7\fbprophet_6fb4ff087aaf40478b71089d7e634e82\setup.py", line 122, in <module>
      setup(
    File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\site-packages\setuptools\__init__.py", line 163, in setup
      return distutils.core.setup(**attrs)
    File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\core.py", line 148, in setup
      dist.run_commands()
    File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\dist.py", line 966, in run_commands
      self.run_command(cmd)
    File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\dist.py", line 985, in run_command
      cmd_obj.run()
    File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\site-packages\wheel\bdist_wheel.py", line 299, in run
      self.run_command('build')
    File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\cmd.py", line 313, in run_command
      self.distribution.run_command(command)
    File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\dist.py", line 985, in run_command
      cmd_obj.run()
    File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\command\build.py", line 135, in run
      self.run_command(cmd_name)
    File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\cmd.py", line 313, in run_command
      self.distribution.run_command(command)
    File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\dist.py", line 985, in run_command
      cmd_obj.run()
    File "C:\Users\matheiss\AppData\Local\Temp\1\pip-install-5dy9pwq7\fbprophet_6fb4ff087aaf40478b71089d7e634e82\setup.py", line 48, in run
      build_models(target_dir)
    File "C:\Users\matheiss\AppData\Local\Temp\1\pip-install-5dy9pwq7\fbprophet_6fb4ff087aaf40478b71089d7e634e82\setup.py", line 38, in build_models
      StanBackendEnum.get_backend_class(backend).build_model(target_dir, MODEL_DIR)
    File "C:\Users\matheiss\AppData\Local\Temp\1\pip-install-5dy9pwq7\fbprophet_6fb4ff087aaf40478b71089d7e634e82\fbprophet\models.py", line 209, in build_model
      import pystan
    File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\site-packages\pystan\__init__.py", line 9, in <module>
      from pystan.api import stanc, stan
    File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\site-packages\pystan\api.py", line 13, in <module>
      import pystan._api  # stanc wrapper
  ImportError: DLL load failed while importing _api: Das angegebene Modul wurde nicht gefunden.
  ----------------------------------------
  ERROR: Failed building wheel for fbprophet
  Running setup.py clean for fbprophet
Failed to build fbprophet
Installing collected packages: fbprophet, ax-platform, attrs, kats
    Running setup.py install for fbprophet ... error
    ERROR: Command errored out with exit status 1:
     command: 'C:\Users\matheiss\Miniconda3\envs\tiresias-kats\python.exe' -u -c 'import io, os, sys, setuptools, tokenize; sys.argv[0] = '"'"'C:\\Users\\matheiss\\AppData\\Local\\Temp\\1\\pip-install-5dy9pwq7\\fbprophet_6fb4ff087aaf40478b71089d7e634e82\\setup.py'"'"'; __file__='"'"'C:\\Users\\matheiss\\AppData\\Local\\Temp\\1\\pip-install-5dy9pwq7\\fbprophet_6fb4ff087aaf40478b71089d7e634e82\\setup.py'"'"';f = getattr(tokenize, '"'"'open'"'"', open)(__file__) if os.path.exists(__file__) else io.StringIO('"'"'from setuptools import setup; setup()'"'"');code = f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, __file__, '"'"'exec'"'"'))' install --record 'C:\Users\matheiss\AppData\Local\Temp\1\pip-record-nj09haiu\install-record.txt' --single-version-externally-managed --compile --install-headers 'C:\Users\matheiss\Miniconda3\envs\tiresias-kats\Include\fbprophet'
         cwd: C:\Users\matheiss\AppData\Local\Temp\1\pip-install-5dy9pwq7\fbprophet_6fb4ff087aaf40478b71089d7e634e82\
    Complete output (44 lines):
    running install
    running build
    running build_py
    creating build
    creating build\lib
    creating build\lib\fbprophet
    creating build\lib\fbprophet\stan_model
    Traceback (most recent call last):
      File "<string>", line 1, in <module>
      File "C:\Users\matheiss\AppData\Local\Temp\1\pip-install-5dy9pwq7\fbprophet_6fb4ff087aaf40478b71089d7e634e82\setup.py", line 122, in <module>
        setup(
      File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\site-packages\setuptools\__init__.py", line 163, in setup
        return distutils.core.setup(**attrs)
      File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\core.py", line 148, in setup
        dist.run_commands()
      File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\dist.py", line 966, in run_commands
        self.run_command(cmd)
      File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\dist.py", line 985, in run_command
        cmd_obj.run()
      File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\site-packages\setuptools\command\install.py", line 61, in run
        return orig.install.run(self)
      File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\command\install.py", line 545, in run
        self.run_command('build')
      File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\cmd.py", line 313, in run_command
        self.distribution.run_command(command)
      File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\dist.py", line 985, in run_command
        cmd_obj.run()
      File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\command\build.py", line 135, in run
        self.run_command(cmd_name)
      File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\cmd.py", line 313, in run_command
        self.distribution.run_command(command)
      File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\distutils\dist.py", line 985, in run_command
        cmd_obj.run()
      File "C:\Users\matheiss\AppData\Local\Temp\1\pip-install-5dy9pwq7\fbprophet_6fb4ff087aaf40478b71089d7e634e82\setup.py", line 48, in run
        build_models(target_dir)
      File "C:\Users\matheiss\AppData\Local\Temp\1\pip-install-5dy9pwq7\fbprophet_6fb4ff087aaf40478b71089d7e634e82\setup.py", line 38, in build_models
        StanBackendEnum.get_backend_class(backend).build_model(target_dir, MODEL_DIR)
      File "C:\Users\matheiss\AppData\Local\Temp\1\pip-install-5dy9pwq7\fbprophet_6fb4ff087aaf40478b71089d7e634e82\fbprophet\models.py", line 209, in build_model
        import pystan
      File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\site-packages\pystan\__init__.py", line 9, in <module>
        from pystan.api import stanc, stan
      File "C:\Users\matheiss\Miniconda3\envs\tiresias-kats\lib\site-packages\pystan\api.py", line 13, in <module>
        import pystan._api  # stanc wrapper
    ImportError: DLL load failed while importing _api: Das angegebene Modul wurde nicht gefunden.
    ----------------------------------------
ERROR: Command errored out with exit status 1: 'C:\Users\matheiss\Miniconda3\envs\tiresias-kats\python.exe' -u -c 'import io, os, sys, setuptools, tokenize; sys.argv[0] = '"'"'C:\\Users\\matheiss\\AppData\\Local\\Temp\\1\\pip-install-5dy9pwq7\\fbprophet_6fb4ff087aaf40478b71089d7e634e82\\setup.py'"'"'; __file__='"'"'C:\\Users\\matheiss\\AppData\\Local\\Temp\\1\\pip-install-5dy9pwq7\\fbprophet_6fb4ff087aaf40478b71089d7e634e82\\setup.py'"'"';f = getattr(tokenize, '"'"'open'"'"', open)(__file__) if os.path.exists(__file__) else io.StringIO('"'"'from setuptools import setup; setup()'"'"');code = f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, __file__, '"'"'exec'"'"'))' install --record 'C:\Users\matheiss\AppData\Local\Temp\1\pip-record-nj09haiu\install-record.txt' --single-version-externally-managed --compile --install-headers 'C:\Users\matheiss\Miniconda3\envs\tiresias-kats\Include\fbprophet' Check the logs for full command output.

I know it is early days, but the package seems to borrow from other solutions already out there, maybe some level of attributions would be good.

fbprophet is no longer the maintained package. It should be changed to prophet, see facebook/prophet#1844 and facebook/prophet@6d81543.

I get the error below while importing CUSUMDetectorTest in Tutorial "Daily Seasonality with Interest Windows" (tkats_202_detection.ipynb)

I solved the issue by copying the CSV file in the subfolder kats/kats/data of my running notebook folder

It looks like a bug in test_detectors.py?

---------------------------------------------------------------------------
FileNotFoundError                         Traceback (most recent call last)
<ipython-input-11-6570ee1979fa> in <module>
----> 1 from kats.tests.test_detectors import CUSUMDetectorTest
      2 
      3 # Here's where we load the seasonal data
      4 np.random.seed(1)
      5 periodicity=24

~\AppData\Local\Continuum\anaconda3\lib\site-packages\kats\tests\test_detectors.py in <module>
    105     multi_data_path = "kats/kats/data/multivariate_anomaly_simulated_data.csv"
    106 
--> 107 data = pd.read_csv(data_path)
    108 data.columns = ["time", "y"]
    109 ts_data = TimeSeriesData(data)
.
.
.
FileNotFoundError: [Errno 2] No such file or directory: 'kats/kats/data/air_passengers.csv'

Hi,

I saw the examples about seasonality_strength in the tutorials.
I went over the source code and it was a bit hard to search until I found these references:
https://stats.stackexchange.com/questions/485012/measuring-strength-of-trend-and-seasonalities-for-time-series-presenting-multi-s
https://otexts.com/fpp2/seasonal-strength.html

I would like some explanation added on how seasonality/trend strength is calculated.
If I can get some resources/guidance on where such document needs to be written I would love to create a PR.

Will it suffice to add some explanation in the example notebooks or should it be documented elsewhere as well?

Hi,

I couldn't reproduce the Holt-Winters code snippet from Kats-201 notebook. Gives NaN as forecast output.

I'm experimenting with BOCPD for regime and trend detection. In my case, the call to the detector function results in the following error when called with choose_priors=True. I'm running the detector on a univariate timeseries of length 280.

changepoints = detector.detector(model=BOCPDModelType.TREND_CHANGE_MODEL, choose_priors=True, lag=7, threshold=0.8)

results in

ValueError: matmul: Input operand 1 does not have enough dimensions (has 0, gufunc core with signature (n?,k),(k,m?)->(n?,m?) requires 1)

Changing choose_priors=False results in no errors.

I'd appreciate your help. Maybe there is something wrong with calling the detector function.

My data (Earth's electromagnetic field components) is typically sampled with a rate 1-100 Hz, with a daily periodicity (following the solar cycle). For the "long period" (1 Hz, i.e. one sample per second) the script detectors/outlier.py makes an error and incorrectly infers that the period is "S" (i.e., a second) and passes "S" to the tsatools module, which can handle only hourly ("H") to annual ("A") periods. Hence, the execution crashes. On the other hand, if I change the line 88 from:
result = seasonal_decompose(original, model=self.decomp)
to
result = seasonal_decompose(original, model=self.decomp, freq=86400)
it runs fine.
Would it be possible to make a provision for the user to supply their own freq value?

Hi,

I am following the steps as requested. I can build models like arima, holter, prophet but I cannot do hyper parameter tuning. I receive error: No frequency information was provided with date index and no frequency could be inferred. Then, I set freq='MS' but I still receive the same error. Any clue?

Thanks,
faye

Hi there,

When running some of the code from the backtesting section of the tutorials as seen below,

`params = ARIMAParams(p=2, d=1, q=1)
ALL_ERRORS = ['mape', 'smape', 'mae', 'mase', 'mse', 'rmse']

backtester_arima = BackTesterSimple(
error_methods=ALL_ERRORS,
data=air_passengers_ts,
params=params,
train_percentage=75,
test_percentage=25,
model_class=ARIMAModel)

backtester_arima.run_backtest()`

there is an issue if you try to replace the air_passengers_ts data with a TimeSeriesData object which contains minute bars that repeat on separate days. For instance, the 'time' column of some new TimeSeriesData object X is of the format 2012-01-02 09:30:00 for row 1 (or rather, x_1) and of the format 2012-01-02 09:31:00 for row 2 (or rather, x_2).

When trying to instantiate a BackTesterSimple object using this data, it succeeds if there is only 1 day's worth of minute bar data. However, if we also have a second day's worth of data with the format 2012-01-03 09:30:00 (x_392) and 2012-01-03 09:31:00 (x_393) and so on... up until the end of the end of the same sequence of minute bars as the previous day, then the code fails with error:

ValueError: No frequency information was provided with date index and no frequency could be inferred.