Automated Time Series Forecasting

Colin Catlin

Last update: Jan 3, 2023

Related tags

Machine Learning machine-learning deep-learning time-series forecasting preprocessing feature-engineering automl autots

Overview

AutoTS

AutoTS is a time series package for Python designed for rapidly deploying high-accuracy forecasts at scale.

There are dozens of forecasting models usable in the sklearn style of .fit() and .predict(). These includes naive, statistical, machine learning, and deep learning models. Additionally, there are over 30 time series specific transforms usable in the sklearn style of .fit(), .transform() and .inverse_transform(). All of these function directly on Pandas Dataframes, without the need for conversion to proprietary objects.

All models support forecasting multivariate (multiple time series) outputs and also support probabilistic (upper/lower bound) forecasts. Most models can readily scale to tens and even hundreds of thousands of input series. Many models also support passing in user-defined exogenous regressors.

These models are all designed for integration in an AutoML feature search which automatically finds the best models, preprocessing, and ensembling for a given dataset through genetic algorithms.

Horizontal and mosaic style ensembles are the flagship ensembling types, allowing each series to receive the most accurate possible models while still maintaining scalability.

A combination of metrics and cross-validation options, the ability to apply subsets and weighting, regressor generation tools, simulation forecastind mode, live datasets, template import and export, plotting, and a collection of data shaping parameters round out the available feature set.

Installation

pip install autots

This includes dependencies for basic models, but additonal packages are required for some models and methods.

Basic Use

Input data for AutoTS is expected to come in either a long or a wide format:

The wide format is a pandas.DataFrame with a pandas.DatetimeIndex and each column a distinct series.
The long format has three columns:
- Date (ideally already in pd.DateTime format)
- Series ID. For a single time series, series_id can be = None.
- Value
For long data, the column name for each of these is passed to .fit() as date_col, id_col, and value_col. No parameters are needed for wide data.

Lower-level functions are only designed for wide style data.

# also load: _hourly, _monthly, _weekly, _yearly, or _live_daily
from autots import AutoTS, load_daily

# sample datasets can be used in either of the long or wide import shapes
long = False
df = load_daily(long=long)

model = AutoTS(
    forecast_length=21,
    frequency='infer',
    prediction_interval=0.9,
    ensemble=None,
    model_list="fast",  # "superfast", "default", "fast_parallel"
    transformer_list="fast",  # "superfast",
    drop_most_recent=1,
    max_generations=4,
    num_validations=2,
    validation_method="backwards"
)
model = model.fit(
    df,
    date_col='datetime' if long else None,
    value_col='value' if long else None,
    id_col='series_id' if long else None,
)

prediction = model.predict()
# plot a sample
prediction.plot(model.df_wide_numeric,
                series=model.df_wide_numeric.columns[0],
                start_date="2019-01-01")
# Print the details of the best model
print(model)

# point forecasts dataframe
forecasts_df = prediction.forecast
# upper and lower forecasts
forecasts_up, forecasts_low = prediction.upper_forecast, prediction.lower_forecast

# accuracy of all tried model results
model_results = model.results()
# and aggregated from cross validation
validation_results = model.results("validation")

The lower-level API, in particular the large section of time series transformers in the scikit-learn style, can also be utilized independently from the AutoML framework.

Check out extended_tutorial.md for a more detailed guide to features!

Also take a look at the production_example.py

Tips for Speed and Large Data:

Use appropriate model lists, especially the predefined lists:
- superfast (simple naive models) and fast (more complex but still faster models, optimized for many series)
- fast_parallel (a combination of fast and parallel) or parallel, given many CPU cores are available
  - n_jobs usually gets pretty close with ='auto' but adjust as necessary for the environment
- see a dict of predefined lists (some defined for internal use) with from autots.models.model_list import model_lists
Use the subset parameter when there are many similar series, subset=100 will often generalize well for tens of thousands of similar series.
- if using subset, passing weights for series will weight subset selection towards higher priority series.
- if limited by RAM, it can be distributed by running multiple instances of AutoTS on different batches of data, having first imported a template pretrained as a starting point for all.
Set model_interrupt=True which passes over the current model when a KeyboardInterrupt ie crtl+c is pressed (although if the interrupt falls between generations it will stop the entire training).
Use the result_file method of .fit() which will save progress after each generation - helpful to save progress if a long training is being done. Use import_results to recover.
While Transformations are pretty fast, setting transformer_max_depth to a lower number (say, 2) will increase speed. Also utilize transformer_list == 'fast' or 'superfast'.
Check out this example of using AutoTS with pandas UDF.
Ensembles are obviously slower to predict because they run many models, 'distance' models 2x slower, and 'simple' models 3x-5x slower.
- ensemble='horizontal-max' with model_list='no_shared_fast' can scale relatively well given many cpu cores because each model is only run on the series it is needed for.
Reducing num_validations and models_to_validate will decrease runtime but may lead to poorer model selections.
For datasets with many records, upsampling (for example, from daily to monthly frequency forecasts) can reduce training time if appropriate.
- this can be done by adjusting frequency and aggfunc but is probably best done before passing data into AutoTS.

How to Contribute:

Give feedback on where you find the documentation confusing
Use AutoTS and...
- Report errors and request features by adding Issues on GitHub
- Posting the top model templates for your data (to help improve the starting templates)
- Feel free to recommend different search grid parameters for your favorite models
And, of course, contributing to the codebase directly on GitHub!

Also known as Project CATS (Catlin's Automated Time Series) hence the logo.

Comments

Exception: Transformer Detrend failed on fit

Hello.
Could you tell me what the message would indicate?

3610 # df = df.replace([np.inf, -np.inf], 0) # .fillna(0) 3611 except Exception as e: -> 3612 raise Exception( 3613 f"Transformer {self.transformations[i]} failed on fit" 3614 ) from e 3615 # df = df.replace([np.inf, -np.inf], 0) # .fillna(0) 3616 return df

Exception: Transformer Detrend failed on fit

opened by r-matsuzaka 12
Their is No Detail about train Data and Test data and no information about [ forecast_length ] in documentation??

Hi Dear. hope you all doing good i want to ask some question about some basic things like how give train data ? how we can give test data ? how we can give target colum ? As i read doc but i never i understand forecast_length what this use for? last question is that as i read in doc catgegorical data must label encode before to feed and is it must to use all int dtype or we can use float dtype columns as well please helpme to answere i would realy appreciate your help. Thanks

opened by faridelya 9

Why is the best result returned by the model not the best according to the resulting dataframe?

Hi,

long = True
# df = load_monthly(long=long)

from autots import AutoTS

metric_weighting = {
    "smape_weighting": 5,
    "mae_weighting": 1,
    "rmse_weighting": 0,
    "containment_weighting": 0,
    "runtime_weighting": 0,
    "spl_weighting": 0,
    "contour_weighting": 0,
}

model = AutoTS(
    forecast_length=30,
    frequency="infer",
    prediction_interval=0.9,
    ensemble=None,
    model_list="superfast",
    transformer_list="fast",
    max_generations=15,
    num_validations=5,
    validation_method="backwards",
    metric_weighting=metric_weighting,
)
model = model.fit(
    df,
    date_col="ds" if long else None,
    value_col="y" if long else None,
    id_col="series" if long else None,
)

prediction = model.predict()

forecasts_df = prediction.forecast
# upper and lower forecasts
forecasts_up, forecasts_low = prediction.upper_forecast, prediction.lower_forecast

# accuracy of all tried model results
model_results = model.results()
validation_results = model.results("validation")

After the run, I get the following Initiated AutoTS object with best model: AverageValueNaive, but then if I sort the validation_results dataframe by Score or smape, the first place is taken by another model, not this one.

Probably, I don't completely understand the way the best model is reported. Thank you in advance!

opened by wilfreddesert 9

SinTrend failed with IndexError('index 1 is out of bounds for axis 0 with size 0')

Hello,

for some reason I get an error that I did not get before. I am using the same code and same data that worked before, but I still get the error.

Anyone else experiencing this issue?

Error: SinTrend failed with IndexError('index 1 is out of bounds for axis 0 with size 0')

Kind regards Luke

opened by LukeBeckerBB 8
self.best_model_name = self.best_model['Model'].iloc[0]

Hi, I am facing difficulty running the model. I have selected some models in a list - best_model. There are 3 of them, and technically model.fit should run all those 3, and give result based on the best one. Sometimes it works, sometimes it throws error with this iloc[0] position statement. It seems it is not able to read the list because I already have 3 algo names there, then how is it not able to read 0th element. Initially i was passing just 1 name, i got same error, then i added 2, now there are 3 algo names, still the same error popping up randomly. What is the root cause of this issue and how can i fix it? Kindly guide.

opened by meetu30 7
Handle Preprocessing for Descending Time Series Data

Detected Unintuitive Behaviour

I ran into a weird error when passing in a dataframe with a descending date index.

ValueError: attempt to get argmax of an empty sequence

Root Cause

autots.tools.shaping.df_cleanup is returning an empty dataframe because the inferred frequency of -1H results in unexpected behaviour when filling missing dates.

Example:

Proposed Change

Let's make sure to sort the time series in ascending order at the beginning of the cleanup for additional robustness and avoid the mystery error message. This also guarantees compatibility with Prophet, which can handle unordered time series data.

opened by adai183 5
Allow use to specify target and features

I cannot understand how autoLS will understand which is my target. For example, passing a wide dataframe how can I understand that that y is my target and x my features for predicting the target?

opened by ghost 5
zero forecast generated using statsmodel

Looks like all the statsmodels model where warnings.simplefilter("ignore", category='ConvergenceWarning') is used are failing and generating zero forecasts

opened by yvadalia 4

Cant' get high VARMAX orders from AutoTS

No matter the value that I set in max_generations arg, the orders of VARMAX that I get are always between 0 and 2. Is there is any way to make AutoTS tries higher orders for VARIMAX models ? I tried various parameters with AutoTS, here is an example of how I create a model and fit it.

model = AutoTS(
    forecast_length=9,
    frequency='M',
    prediction_interval=0.9,
    model_list=["VARMAX"],
    ensemble="all",
    max_generations=100,
    num_validations=2,
    validation_method="backwards",
    random_seed=7,
    models_to_validate = 0.2,
    transformer_list="all",
    transformer_max_depth=8
)
# VARMA_train_data is a wide data set of 6 columns and 54 monthly observasions
model.fit(VARMA_train_data)

opened by Abdelgha-4 4

Question: Optimal max_generations

I wonder how to set the generation number.
Is there any optimal number?

Also, each generation tries only one model randomly?
Then I should choose more than the number of models available like 40?

opened by r-matsuzaka 2
When would be available early stopping option?

Hello, again. I found the comment, It's called 'max' because someday there will be an auto early stopping option, but for now this is just the exact number of generations to run. at https://github.com/winedarksea/AutoTS/blob/52487e5ea5e3ce0530e1096e239655786227de3f/autots/evaluator/auto_ts.py#L54

Do you plan to implement this feature?

opened by r-matsuzaka 2
Question: Multi-Objective Forecasting

Problem: A dataset with multiple data columns that may or may not be temporally coupled, and also multiple output sets. e.g. market sectors and wellbeing index as columns, see if one index is tied to the rest of the indices.

opened by BrandonKMLee 7
Evaluation Metrics are missing and all models have failed, by an error in template or metrics
Describe the bug I tried to use DNN models presented in the package such as GluonTS, pytorch-forecasting and tensorflowSTS. I have installed all required packages as in the installation guide (My environment can be viewed below). The error I got from the fit() function read as:

"Evaluation Metrics are missing and all models have failed, by an error in template or metrics. There are many possible causes for this, bad parameters, environment, or an unreported bug. Usually this means you are missing required packages for the models like fbprophet or gluonts, or that the models in model_list are inappropriate for your data. A new starting template may also help".

To Reproduce When I run the following code:

from autots import AutoTS model_list = ['TensorflowSTS'] model = AutoTS( forecast_length=7, frequency='D', prediction_interval=0.9, metric_weighting=metric_weighting, ensemble='horizontal', holiday_country='UK', model_list=model_list, drop_most_recent=0, max_generations=1, num_validations=3, validation_method="backwards", no_negatives=True, verbose=-2, n_jobs = 72 ) model = model.fit(train_df)

Expected behavior Stack trace:

--------------------------------------------------------------------------- KeyError Traceback (most recent call last) File ~/anaconda3/envs/forecast/lib/python3.10/site-packages/pandas/core/indexes/base.py:3621, in Index.get_loc(self, key, method, tolerance) 3620 try: -> 3621 return self._engine.get_loc(casted_key) 3622 except KeyError as err: File ~/anaconda3/envs/forecast/lib/python3.10/site-packages/pandas/_libs/index.pyx:136, in pandas._libs.index.IndexEngine.get_loc() File ~/anaconda3/envs/forecast/lib/python3.10/site-packages/pandas/_libs/index.pyx:163, in pandas._libs.index.IndexEngine.get_loc() File pandas/_libs/hashtable_class_helper.pxi:5198, in pandas._libs.hashtable.PyObjectHashTable.get_item() File pandas/_libs/hashtable_class_helper.pxi:5206, in pandas._libs.hashtable.PyObjectHashTable.get_item() KeyError: 'smape' The above exception was the direct cause of the following exception: KeyError Traceback (most recent call last) File ~/anaconda3/envs/forecast/lib/python3.10/site-packages/autots/evaluator/auto_model.py:2031, in generate_score(model_results, metric_weighting, prediction_interval) 2030 # not sure why there are negative SMAPE values, but make sure they get dealt with -> 2031 if model_results['smape'].min() < 0: 2032 model_results['smape'] = model_results['smape'].where( 2033 model_results['smape'] >= 0, model_results['smape'].max() 2034 ) File ~/anaconda3/envs/forecast/lib/python3.10/site-packages/pandas/core/frame.py:3505, in DataFrame.__getitem__(self, key) 3504 return self._getitem_multilevel(key) -> 3505 indexer = self.columns.get_loc(key) 3506 if is_integer(indexer): File ~/anaconda3/envs/forecast/lib/python3.10/site-packages/pandas/core/indexes/base.py:3623, in Index.get_loc(self, key, method, tolerance) 3622 except KeyError as err: -> 3623 raise KeyError(key) from err 3624 except TypeError: 3625 # If we have a listlike key, _check_indexing_error will raise 3626 # InvalidIndexError. Otherwise we fall through and re-raise 3627 # the TypeError. KeyError: 'smape' During handling of the above exception, another exception occurred: KeyError Traceback (most recent call last) File <timed exec>:44, in <module> File ~/anaconda3/envs/forecast/lib/python3.10/site-packages/autots/evaluator/auto_ts.py:791, in AutoTS.fit(self, df, date_col, value_col, id_col, future_regressor, weights, result_file, grouping_ids, validation_indexes) 789 # capture the data from the lower level results 790 self.initial_results = self.initial_results.concat(template_result) --> 791 self.initial_results.model_results['Score'] = generate_score( 792 self.initial_results.model_results, 793 metric_weighting=metric_weighting, 794 prediction_interval=prediction_interval, 795 ) 796 if result_file is not None: 797 self.initial_results.save(result_file) File ~/anaconda3/envs/forecast/lib/python3.10/site-packages/autots/evaluator/auto_model.py:2126, in generate_score(model_results, metric_weighting, prediction_interval) 2123 overall_score = overall_score + (containment_score * containment_weighting) 2125 except Exception as e: -> 2126 raise KeyError( 2127 f"""Evaluation Metrics are missing and all models have failed, by an error in template or metrics. 2128 There are many possible causes for this, bad parameters, environment, or an unreported bug. 2129 Usually this means you are missing required packages for the models like fbprophet or gluonts, 2130 or that the models in model_list are inappropriate for your data. 2131 A new starting template may also help. {repr(e)}""" 2132 ) 2134 return overall_score.astype(float)

Desktop (please complete the following information):

OS: Ubuntu 20.04

Package Versions 0.4.2

``

Name Version Build Channel

_libgcc_mutex 0.1 conda_forge conda-forge _openmp_mutex 4.5 2_gnu conda-forge abseil-cpp 20210324.2 h9c3ff4c_0 conda-forge absl-py 1.1.0 pyhd8ed1ab_0 conda-forge aiohttp 3.8.1 py310h5764c6d_1 conda-forge aiosignal 1.2.0 pyhd8ed1ab_0 conda-forge alembic 1.8.0 pyhd8ed1ab_0 conda-forge appdirs 1.4.4 pypi_0 pypi arch 5.3.1 pypi_0 pypi argon2-cffi 21.3.0 pypi_0 pypi argon2-cffi-bindings 21.2.0 pypi_0 pypi asttokens 2.0.5 pypi_0 pypi astunparse 1.6.3 pyhd8ed1ab_0 conda-forge async-timeout 4.0.2 pyhd8ed1ab_0 conda-forge attrs 21.4.0 pyhd8ed1ab_0 conda-forge autopage 0.5.1 pyhd8ed1ab_0 conda-forge autots 0.4.2 pyhd8ed1ab_0 conda-forge backcall 0.2.0 pypi_0 pypi backports 1.0 py_2 conda-forge backports.functools_lru_cache 1.6.4 pyhd8ed1ab_0 conda-forge beautifulsoup4 4.11.1 pypi_0 pypi blas 1.0 mkl conda-forge bleach 5.0.0 pypi_0 pypi blinker 1.4 py_1 conda-forge brotli 1.0.9 h166bdaf_7 conda-forge brotli-bin 1.0.9 h166bdaf_7 conda-forge brotlipy 0.7.0 py310h5764c6d_1004 conda-forge bzip2 1.0.8 h7f98852_4 conda-forge c-ares 1.18.1 h7f98852_0 conda-forge ca-certificates 2022.6.15 ha878542_0 conda-forge cached-property 1.5.2 hd8ed1ab_1 conda-forge cached_property 1.5.2 pyha770c72_1 conda-forge cachetools 5.0.0 pyhd8ed1ab_0 conda-forge certifi 2022.6.15 py310hff52083_0 conda-forge cffi 1.15.0 py310h0fdd8cc_0 conda-forge charset-normalizer 2.0.12 pyhd8ed1ab_0 conda-forge click 8.1.3 py310hff52083_0 conda-forge cliff 3.10.1 pyhd8ed1ab_0 conda-forge clikit 0.6.2 pypi_0 pypi cloudpickle 2.1.0 pyhd8ed1ab_0 conda-forge cmaes 0.8.2 pyh44b312d_0 conda-forge cmd2 2.3.3 py310hff52083_1 conda-forge cmdstanpy 0.9.5 pypi_0 pypi colorama 0.4.5 pyhd8ed1ab_0 conda-forge colorlog 6.6.0 py310hff52083_1 conda-forge convertdate 2.4.0 pypi_0 pypi crashtest 0.3.1 pypi_0 pypi cryptography 37.0.1 py310h9ce1e76_0
cudatoolkit 11.3.1 h9edb442_10 conda-forge cycler 0.11.0 pyhd8ed1ab_0 conda-forge cython 0.29.30 pypi_0 pypi dask 2022.6.1 pypi_0 pypi debugpy 1.6.0 pypi_0 pypi decorator 5.1.1 pypi_0 pypi defusedxml 0.7.1 pypi_0 pypi dill 0.3.5.1 pypi_0 pypi distributed 2022.6.1 pypi_0 pypi entrypoints 0.4 pypi_0 pypi ephem 4.1.3 pypi_0 pypi executing 0.8.3 pypi_0 pypi fastjsonschema 2.15.3 pypi_0 pypi fonttools 4.33.3 pypi_0 pypi freetype 2.10.4 h0708190_1 conda-forge frozenlist 1.3.0 py310h5764c6d_1 conda-forge fsspec 2022.5.0 pyhd8ed1ab_0 conda-forge future 0.18.2 py310hff52083_5 conda-forge gast 0.5.3 pyhd8ed1ab_0 conda-forge giflib 5.2.1 h36c2ea0_2 conda-forge gluonts 0.10.1 pypi_0 pypi google-auth 2.9.0 pyh6c4a22f_0 conda-forge google-auth-oauthlib 0.4.1 py_2 conda-forge google-pasta 0.2.0 pyh8c360ce_0 conda-forge greenlet 1.1.2 py310hd8f1fbe_2 conda-forge greykite 0.3.0 pypi_0 pypi grpc-cpp 1.45.2 h3b8df00_4 conda-forge grpcio 1.45.0 py310h44b9e0c_0 conda-forge h5py 3.7.0 nompi_py310h06dffec_100 conda-forge hdf5 1.12.1 nompi_h2386368_104 conda-forge heapdict 1.0.1 pypi_0 pypi hijri-converter 2.2.4 pypi_0 pypi holidays 0.14.2 pypi_0 pypi holidays-ext 0.0.7 pypi_0 pypi httpstan 4.7.2 pypi_0 pypi icu 70.1 h27087fc_0 conda-forge idna 3.3 pyhd8ed1ab_0 conda-forge importlib-metadata 4.11.4 py310hff52083_0 conda-forge importlib_resources 5.8.0 pyhd8ed1ab_0 conda-forge intel-openmp 2022.0.1 h06a4308_3633
ipykernel 6.15.0 pypi_0 pypi ipython 8.4.0 pypi_0 pypi ipython-genutils 0.2.0 pypi_0 pypi jedi 0.18.1 pypi_0 pypi jinja2 3.1.2 pypi_0 pypi joblib 1.1.0 pyhd8ed1ab_0 conda-forge jpeg 9e h166bdaf_2 conda-forge jsonschema 4.6.0 pypi_0 pypi jupyter-client 7.3.4 pypi_0 pypi jupyter-console 6.4.4 pypi_0 pypi jupyter-core 4.10.0 pypi_0 pypi jupyter-http-over-ws 0.0.8 pypi_0 pypi jupyterlab-pygments 0.2.2 pypi_0 pypi jupyterlab-widgets 1.1.1 pypi_0 pypi keras 2.8.0 pyhd8ed1ab_0 conda-forge keras-preprocessing 1.1.2 pyhd8ed1ab_0 conda-forge keyutils 1.6.1 h166bdaf_0 conda-forge kiwisolver 1.4.3 py310hbf28c38_0 conda-forge korean-lunar-calendar 0.2.1 pypi_0 pypi krb5 1.19.3 h3790be6_0 conda-forge lcms2 2.12 hddcbb42_0 conda-forge ld_impl_linux-64 2.36.1 hea4e1c9_2 conda-forge lerc 3.0 h9c3ff4c_0 conda-forge libblas 3.9.0 14_linux64_mkl conda-forge libbrotlicommon 1.0.9 h166bdaf_7 conda-forge libbrotlidec 1.0.9 h166bdaf_7 conda-forge libbrotlienc 1.0.9 h166bdaf_7 conda-forge libcblas 3.9.0 14_linux64_mkl conda-forge libcurl 7.83.1 h7bff187_0 conda-forge libdeflate 1.12 h166bdaf_0 conda-forge libedit 3.1.20191231 he28a2e2_2 conda-forge libev 4.33 h516909a_1 conda-forge libffi 3.4.2 h7f98852_5 conda-forge libgcc-ng 12.1.0 h8d9b700_16 conda-forge libgfortran-ng 12.1.0 h69a702a_16 conda-forge libgfortran5 12.1.0 hdcd56e2_16 conda-forge libgomp 12.1.0 h8d9b700_16 conda-forge liblapack 3.9.0 14_linux64_mkl conda-forge libnghttp2 1.47.0 h727a467_0 conda-forge libnsl 2.0.0 h7f98852_0 conda-forge libopenblas 0.3.20 pthreads_h78a6416_0 conda-forge libpng 1.6.37 h753d276_3 conda-forge libprotobuf 3.20.1 h6239696_0 conda-forge libssh2 1.10.0 ha56f1ee_2 conda-forge libstdcxx-ng 12.1.0 ha89aaad_16 conda-forge libtiff 4.4.0 hc85c160_1 conda-forge libuuid 2.32.1 h7f98852_1000 conda-forge libwebp 1.2.2 h3452ae3_0 conda-forge libwebp-base 1.2.2 h7f98852_1 conda-forge libxcb 1.13 h7f98852_1004 conda-forge libzlib 1.2.12 h166bdaf_1 conda-forge lightgbm 3.3.2 py310h122e73d_0 conda-forge locket 1.0.0 pypi_0 pypi lunarcalendar 0.0.9 pypi_0 pypi lz4-c 1.9.3 h9c3ff4c_1 conda-forge mako 1.2.1 pyhd8ed1ab_0 conda-forge markdown 3.3.7 pyhd8ed1ab_0 conda-forge markupsafe 2.1.1 py310h5764c6d_1 conda-forge marshmallow 3.16.0 pypi_0 pypi matplotlib-base 3.5.2 py310h5701ce4_0 conda-forge matplotlib-inline 0.1.3 pypi_0 pypi mistune 0.8.4 pypi_0 pypi mkl 2022.0.1 h06a4308_117
msgpack 1.0.4 pypi_0 pypi multidict 6.0.2 py310h5764c6d_1 conda-forge munkres 1.1.4 pyh9f0ad1d_0 conda-forge mxnet-cu112 1.9.1 pypi_0 pypi nbclient 0.6.4 pypi_0 pypi nbconvert 6.5.0 pypi_0 pypi nbformat 5.4.0 pypi_0 pypi ncurses 6.3 h27087fc_1 conda-forge nest-asyncio 1.5.5 pypi_0 pypi notebook 6.4.12 pypi_0 pypi numpy 1.23.0 py310h53a5b5f_0 conda-forge oauthlib 3.2.0 pyhd8ed1ab_0 conda-forge openjpeg 2.4.0 hb52868f_1 conda-forge openssl 1.1.1q h166bdaf_0 conda-forge opt_einsum 3.3.0 pyhd8ed1ab_1 conda-forge optuna 2.10.1 pyhd8ed1ab_0 conda-forge packaging 21.3 pyhd8ed1ab_0 conda-forge pandas 1.4.3 py310h769672d_0 conda-forge pandocfilters 1.5.0 pypi_0 pypi parso 0.8.3 pypi_0 pypi partd 1.2.0 pypi_0 pypi pastel 0.2.1 pypi_0 pypi patsy 0.5.2 pyhd8ed1ab_0 conda-forge pbr 5.9.0 pyhd8ed1ab_0 conda-forge pexpect 4.8.0 pypi_0 pypi pickleshare 0.7.5 pypi_0 pypi pillow 9.1.1 pypi_0 pypi pip 22.1.2 pyhd8ed1ab_0 conda-forge pmdarima 1.8.5 pypi_0 pypi prettytable 3.3.0 pypi_0 pypi prometheus-client 0.14.1 pypi_0 pypi prompt-toolkit 3.0.29 pypi_0 pypi property-cached 1.6.4 pypi_0 pypi prophet 1.1 pypi_0 pypi protobuf 3.20.1 py310hd8f1fbe_0 conda-forge psutil 5.9.1 pypi_0 pypi pthread-stubs 0.4 h36c2ea0_1001 conda-forge ptyprocess 0.7.0 pypi_0 pypi pure-eval 0.2.2 pypi_0 pypi pyasn1 0.4.8 py_0 conda-forge pyasn1-modules 0.2.7 py_0 conda-forge pycparser 2.21 pyhd8ed1ab_0 conda-forge pydantic 1.9.1 pypi_0 pypi pydeprecate 0.3.2 pyhd8ed1ab_0 conda-forge pygments 2.12.0 pypi_0 pypi pyjwt 2.4.0 pyhd8ed1ab_0 conda-forge pylev 1.4.0 pypi_0 pypi pymeeus 0.5.11 pypi_0 pypi pyopenssl 22.0.0 pyhd8ed1ab_0 conda-forge pyparsing 3.0.9 pyhd8ed1ab_0 conda-forge pyperclip 1.8.2 pyhd8ed1ab_2 conda-forge pyrsistent 0.18.1 pypi_0 pypi pysimdjson 3.2.0 pypi_0 pypi pysocks 1.7.1 py310hff52083_5 conda-forge pystan 3.4.0 pypi_0 pypi python 3.10.5 h582c2e5_0_cpython conda-forge python-dateutil 2.8.2 pyhd8ed1ab_0 conda-forge python-flatbuffers 2.0 pyhd8ed1ab_0 conda-forge python_abi 3.10 2_cp310 conda-forge pytorch 1.12.0 py3.10_cuda11.3_cudnn8.3.2_0 pytorch pytorch-forecasting 0.10.2 pyhd8ed1ab_0 conda-forge pytorch-lightning 1.6.4 pyhd8ed1ab_0 conda-forge pytorch-mutex 1.0 cuda pytorch pytz 2022.1 pyhd8ed1ab_0 conda-forge pyu2f 0.1.5 pyhd8ed1ab_0 conda-forge pyyaml 5.4.1 pypi_0 pypi pyzmq 23.2.0 pypi_0 pypi qtconsole 5.3.1 pypi_0 pypi qtpy 2.1.0 pypi_0 pypi re2 2022.06.01 h27087fc_0 conda-forge readline 8.1.2 h0f457ee_0 conda-forge requests 2.28.0 pyhd8ed1ab_0 conda-forge requests-oauthlib 1.3.1 pyhd8ed1ab_0 conda-forge rsa 4.8 pyhd8ed1ab_0 conda-forge scikit-learn 1.1.1 py310hffb9edd_0 conda-forge scipy 1.8.1 py310h7612f91_0 conda-forge seaborn 0.11.2 pypi_0 pypi send2trash 1.8.0 pypi_0 pypi setuptools 59.5.0 py310hff52083_0 conda-forge setuptools-git 1.2 pypi_0 pypi six 1.16.0 pyh6c4a22f_0 conda-forge snappy 1.1.9 hbd366e4_1 conda-forge sortedcontainers 2.4.0 pypi_0 pypi soupsieve 2.3.2.post1 pypi_0 pypi sqlalchemy 1.4.39 py310h5764c6d_0 conda-forge sqlite 3.38.5 h4ff8645_0 conda-forge stack-data 0.3.0 pypi_0 pypi statsmodels 0.13.2 py310hde88566_0 conda-forge stevedore 3.5.0 py310hff52083_3 conda-forge tblib 1.7.0 pypi_0 pypi tensorboard 2.8.0 pyhd8ed1ab_1 conda-forge tensorboard-data-server 0.6.0 py310h597c629_2 conda-forge tensorboard-plugin-wit 1.8.1 pyhd8ed1ab_0 conda-forge tensorflow 2.8.1 cpu_py310hd1aba9c_0 conda-forge tensorflow-base 2.8.1 cpu_py310h17449b8_0 conda-forge tensorflow-estimator 2.8.1 cpu_py310had6d012_0 conda-forge termcolor 1.1.0 pyhd8ed1ab_3 conda-forge terminado 0.15.0 pypi_0 pypi threadpoolctl 3.1.0 pyh8a188c0_0 conda-forge tinycss2 1.1.1 pypi_0 pypi tk 8.6.12 h27826a3_0 conda-forge toolz 0.11.2 pypi_0 pypi torchmetrics 0.9.2 pyhd8ed1ab_0 conda-forge tornado 6.1 pypi_0 pypi tqdm 4.64.0 pyhd8ed1ab_0 conda-forge traitlets 5.3.0 pypi_0 pypi typing-extensions 4.3.0 hd8ed1ab_0 conda-forge typing_extensions 4.3.0 pyha770c72_0 conda-forge tzdata 2022a h191b570_0 conda-forge ujson 5.3.0 pypi_0 pypi unicodedata2 14.0.0 py310h5764c6d_1 conda-forge urllib3 1.26.9 pyhd8ed1ab_0 conda-forge wcwidth 0.2.5 pyh9f0ad1d_2 conda-forge webargs 8.1.0 pypi_0 pypi webencodings 0.5.1 pypi_0 pypi werkzeug 2.1.2 pyhd8ed1ab_1 conda-forge wheel 0.37.1 pyhd8ed1ab_0 conda-forge widgetsnbextension 3.6.1 pypi_0 pypi wrapt 1.14.1 py310h5764c6d_0 conda-forge xgboost 1.6.1 pypi_0 pypi xlrd 2.0.1 pypi_0 pypi xorg-libxau 1.0.9 h7f98852_0 conda-forge xorg-libxdmcp 1.1.3 h7f98852_0 conda-forge xz 5.2.5 h516909a_1 conda-forge yaml 0.2.5 h7f98852_2 conda-forge yarl 1.7.2 py310h5764c6d_2 conda-forge zict 2.2.0 pypi_0 pypi zipp 3.8.0 pyhd8ed1ab_0 conda-forge zlib 1.2.12 h166bdaf_1 conda-forge zstd 1.5.2 h8a70e8d_2 conda-forge ``
opened by JohanObluda 1

Releases(0.5.3)

0.5.3(Dec 23, 2022)
0.5.3 🌍 🌍 🌍

robustness changes to generate_score_per_series for horizontal ensembles

added generation_timeout to allow stopping based on time

added ability to specify anomaly model (ie 'zscore') to AnomalyDetector.get_new_params(method='zscore')

fixed a bug in Cassandra with future_impacts and future_regressor

Source code(tar.gz)
Source code(zip)
AutoTS-0.5.3-py3-none-any.whl(658.72 KB)
AutoTS-0.5.3.tar.gz(636.39 KB)
0.5.2(Dec 13, 2022)
0.5.2 🌎 🌎 🌎

allow transformer_max_depth == 0

linear_tree option to LightGBM models

additional linear model options to Cassandra

Source code(tar.gz)
Source code(zip)
AutoTS-0.5.2-py3-none-any.whl(658.22 KB)
AutoTS-0.5.2.tar.gz(635.88 KB)
0.5.1(Nov 15, 2022)
0.5.1 :mage_man: :mage_man: :mage_man:

add LocalLinearTrend transformer

improved ScipyFilter focusing on Butter and Savitzky–Golay filters

update to AutoTS().back_forecast including breaking change of column arg renamed to series

add KalmanSmoothing transformer

add KalmanStateSpace model (Kalman Filter + 'any' state space models)

AlignLastValue no longer applied to upper/lower forecast bounds

modified regression impact in HolidayTransformer to weighted least squares, moved existing to 'datepart_regression'

added Cassandra model

bug fix for Categorical dateparts with 1 starts

updated load_daily to Wikimedia page views

added dwae metric

added MetricMotif model

added 'common_fourier' datepart method

added SeasonalMotif model

added 'seasonal' validation

bug fix for cffilter with univariate input

Source code(tar.gz)
Source code(zip)
AutoTS-0.5.1-py3-none-any.whl(656.59 KB)
AutoTS-0.5.1.tar.gz(1.66 MB)
0.5.0(Aug 24, 2022)
0.5.0 :space_invader: :space_invader: :space_invader:

added AnomalyDetector

added HolidayDetector

added observation_end and Wikipedia data to load_live_daily

added binarized versions of datepart method (should have done ages ago!)

addded RRVAR, MAR, TMF, LATC models

added AlignLastValue transformer

added plot_horizontal_model_count and fixed an error in horizontal generation plot

adjusted TotalRuntime to higher precision, and no longer + 1

added subsidiary transformer for cleaning in Detrend and Datepart detrend Transformers

sped up SinTrend transformer

new AnomalyRemoval transformer

added HolidayTransformer

added auto holidays to Prophet

added get_new_params method to AutoTS class

more holidays options to create_regressor

Source code(tar.gz)
Source code(zip)
AutoTS-0.5.0-py3-none-any.whl(517.32 KB)
AutoTS-0.5.0.tar.gz(1.47 MB)
0.4.2(Jun 20, 2022)
Latest :space_invader: :space_invader: :space_invader:

plot_horizontal_per_generation and horizontal_per_generation added

model_list now accepts a dictionary of probabilities, however this only affects new Random Templates

:seedling: improved the genetic algorithm for new model generation

minor improvement to generate_score_per_series for handling very small ~e-20 errors

added PytorchForecasting to available models

Johansen Cointegration transformer

BTCD transformer

Johansen and BTCD as Regression features

fixed bug in plot_horizontal()

added ARCH to available models

changed the sklearn models used by UnivariateRegression by default and returned to default model_list

fixed a bug in KerasRNN

Source code(tar.gz)
Source code(zip)
AutoTS-0.4.2-py3-none-any.whl(475.47 KB)
AutoTS-0.4.2.tar.gz(456.93 KB)
0.4.1(May 16, 2022)
Latest :fireworks: :fireworks:

Replaced ZeroesNaive model with ConstantNaive

updated the General starting template

added 'window' to AverageValueNaive

added :octocat: load_artificial sample dataset

fixed bug in plot_horizontal where not handling negative series

major update to Constraint functionality

GluonTS is no longer part of the default model list (faster tests this way) but is now part of 'best'

horizontal models_to_use for Mosaic ensembles

added some intel optimizations to sklearn code if scikit-learn-intelex installed

fixed a :bug: in rolling_x_regressor where datepart method wasn't actually getting appended

sped up rolling_x_regressor by reducing concats

added current_model_file as an option for additional debugging information

updated PCA and FastICA to be more flexible on n_components

fixed a bug where unpack_ensemble_models with keep_ensemble=False was still keeping nested ensembles

impute speed optimizations for fill_mean, use of nan check so runs faster if no nan present

optimizations to metrics including faster if no NaN present

:heavy_plus_sign: faster percentile function, used in transforms, basics

also added nan_checks to switch between numpy na and numpy non-na quantiles/medians

made fake_date_fill nan method vectorized

slightly adjusted the upper/lower forecast method for LastValueNaive :crystal_ball:

updated sliding_window_view to allow Motifs to run on Numpy < 1.20 and also for faster WindowRegression

updated regressor_used and used_regressor_check so they should be more reliably filled.

changed behavior where import_templates would fail if model_list not satisified. Still fails, but now only for the "only" import option

added :chart_with_upwards_trend: when a model in validation is the best so far that round

addition of "auto" and "max" num_validations with auto set as default

addition of new metrics: mqae, oda, maxe

Source code(tar.gz)
Source code(zip)
AutoTS-0.4.1-py3-none-any.whl(459.54 KB)
AutoTS-0.4.1.tar.gz(443.17 KB)
0.4.0(Feb 7, 2022)
Latest

Note: the plan is to replace ZeroesNaive model with ConstantNaive in a future release

fix bug where score was failing to generate if made_weighting > 0 and forecast_length = 1

made MADE functional on forecast_length=1 if df_train is provided

SMAPE now in repr of fit AutoTS class

contour now works on forecast_length = 1

Added NeuralProphet model

made the probabilistic modeling of MultivariateRegression a parameter which only occurs when 'deep' mode is active (too slow)

added more params to pass through to Prophet

add phi damping to Detrend transformer

add window slice to Detrend transformer

added 'simple_2" datepart method

added package to conda-forge

preclean method added to AutoTS

added median and midhinge point_methods to BestN Ensembles

added additional model selections to 'simple' and 'subsample' ensemble

switch LightGBM back to MultiOutputRegressor from RegressorChain for speed (with LightGBMRegressorChain replacing)

removed top-level datasets requests dependency

Added EWMAFilter

improved NaN in forecast check

added fail_on_forecast_nan (bool) to AutoTS.predict and model_forecast, if False, can now allow forecasts with NaN to be returned

add return_model to model_forecast for model and transformer

fixed bug where Detrend failing with non-datetime index

improved error handling in Transformers to explicitly reference which failed

added random.seed() setting in AutoTS which actually seems to standardize the runs

sped up assembling/concat of horizontal ensembles for large numbers of series

added polynomial_degree to date_part (and ~Transformer and ~Regression)

updated infer_frequency and utilized in model base class

added additional datasets (analytics.gov and severe weather) to load_lively_daily and modified pytrends load

added rps to metrics (although no plans to build it into evaluation)

added 'Ridge' and 'GaussianProcessRegressor' as model options for Regressions

enforcing consistency on inner n_jobs with MultiOutputRegressor

add DynamicFactorMQ model from Statsmodels

added plot_per_series_smape and list_failed_model_types to output more run information from AutoTS class

increased number of best per series models added to models to validate (models to validate has become more of a baseline than a firm number)

finally transitioned ensemble parameter fully to a list from the original comma-sep string list

MLE and iMLE logarithmic metrics for targeting under- and over- estimation

MAGE metric for error on rollup forecasts

Mosaic ensembles now include a metric_weighting variation including MAE, RMSE and SPL weighting (abs error, square error, pl error) (unscaled)

minor but noticeable speedups to TemplateWizard and inferred_normal functions

added EventRiskForecast for determing risk of exceeding limits (should be considered in beta for now)

back_forecast now has tail/eval_periods configured

changed behavior of import_template, by default simple ensembles are unpacked but no longer included in template unless include_ensemble is True.

Source code(tar.gz)
Source code(zip)
AutoTS-0.4.0-py3-none-any.whl(449.97 KB)
AutoTS-0.4.0.tar.gz(430.91 KB)
0.3.12(Dec 5, 2021)
Latest

add MADE error metric (consider this beta, it may change)

"end_generation" option to model_interrupt

statsmodels warning adjustment (warnings now print at verbose = 2)

add modifier to cpu_count and use with model_forecast auto_model

Source code(tar.gz)
Source code(zip)
AutoTS-0.3.12-py3-none-any.whl(431.80 KB)
AutoTS-0.3.12.tar.gz(413.77 KB)
0.3.11(Nov 29, 2021)
Latest

added Benchmark function

made Prophet a bit more robust for joblib

motifsimulation bug fix

simulation forecasting mode

deep param mode (not yet utilized widely)

Source code(tar.gz)
Source code(zip)
AutoTS-0.3.11-py3-none-any.whl(430.87 KB)
AutoTS-0.3.11.tar.gz(412.91 KB)
0.3.10(Nov 4, 2021)
Latest

BestN ensembles now support weighting model components

cluster-based and generate_score_per_series-based 'simple' ensembles

'univariate' model_list added

similarity and custom cross validation now set initial evaluation segment

validation_test_indexes and train now include initial eval segment

'subsample' ensemble expansion of 'simple'

added Theta model from statsmodels

added ARDL model from statsmodels

expanded UnobservedComponents functionality, although it still fails on some params for unknown reasons

fixed bug in AutoTS.predict() where it was breaking regressors in some cases

transition from [] to None as default for no future_regressor

enforce more extensive failing if regression_type==User and no regressor passed

fixed regressor handling in DatepartRegression

Source code(tar.gz)
Source code(zip)
AutoTS-0.3.10-py3-none-any.whl(427.43 KB)
AutoTS-0.3.10.tar.gz(409.74 KB)
0.3.9(Oct 31, 2021)
Latest

update validation template creation for horizontal ensembles

made MultivariateRegression probabilistic

fixed bug where weighting didn't take floats

pushed the evaluate options from a separate function to part of the PredictionObject

added 'custom' validation option

added "similarity" validation option

SectionalMotif model added

window functions grouped in module

fixed bugs in holiday_flag

holiday_flag now has holiday categorical encoding option and works better on sub-daily data

create_regressor handle categorical features

'superfast' transformer_dict now adjusts fillna methods as well

optimizing metric calculation runtimes (feel the speed of 500 µs savings!)

Source code(tar.gz)
Source code(zip)
AutoTS-0.3.9-py3-none-any.whl(423.14 KB)
AutoTS-0.3.9.tar.gz(405.54 KB)
0.3.8(Oct 17, 2021)
Latest

add Transformer model to sklearn DNN models

expanded and tuned KerasRNN model options

added param space for RandomForest, ExtraTrees, Poisson, and RANSAC regressions

removed Tensorflow models from UnivariateRegression as it can cause a crash with GPU training

added create_regressor function

two new impute methods (KNNImputer, IterativeImputerExtraTrees), but only with "all" transformers

deletion of old TSFresh model, which was horribly slow and not going to get any faster

optimizing scalability by tuning transformer and imputation defaults

MultivariateRegression model (RollingRegression but 1d to models)

fix for generate_score_per_series bug with all zeroes series

bug fix for where horizontal ensembles failed if series_ids/column names were integers

Source code(tar.gz)
Source code(zip)
AutoTS-0.3.8-py3-none-any.whl(417.56 KB)
AutoTS-0.3.8.tar.gz(400.65 KB)
0.3.7(Oct 4, 2021)
Latest

bug fix in fake_date imputation

bug fix in Round

make SinTrend fail if it fails on all series (may revert this)

load_linear and load_sine artificial datasets

new NVAR model based on https://github.com/quantinfo/ng-rc-paper-code/

tuning retrieve_regressor to allow it to better work with multioutput and univariate

expand GluonTS models included

GluonTS now works on univariate inputs

GluonTS now works with regressors

fixed bug where model_count wrong for mosaic ensembles

fixed bug in VECM that meant it didn't couldn't utilize future_regressor

Source code(tar.gz)
Source code(zip)
AutoTS-0.3.7-py3-none-any.whl(410.80 KB)
AutoTS-0.3.7.tar.gz(395.64 KB)
0.3.6(Sep 14, 2021)
Latest

back_forecast for forecast on training data

Mosaic ensembles can now be used beyond training forecast_length and for shorter lengths too

best_model_name, best_model_params, and best_model_transformation_params AutoTS attributes now available

mean, median, and ffill NaN now handle fully NaN series by returning 0.

fixed bug that was causing mosaic generalization to fail if ffill/bfill handled all missing values

STLFilter and HPFilter and convolution_filter Transformers added

Source code(tar.gz)
Source code(zip)
AutoTS-0.3.6-py3-none-any.whl(406.10 KB)
AutoTS-0.3.6.tar.gz(391.08 KB)
0.3.5(Aug 30, 2021)
Latest

New Transfromer ScipyFilter

New models Univariate and MultivariateMotif

'midhinge' and "weighted_mean" to AverageValueNaive

Add passing regressors to WindowRegression and made more efficient window generation

more plotting methods: plot_horizontal_transformers

for most -Regression type models, model_params is now treated as kwargs and can accept any args for that model

ExtraTrees and RadiusRegressor to -Regression type models

bug fix in generate_score_per_series

'Generation' now tracked in results table, plus plotting method for generation loss

Source code(tar.gz)
Source code(zip)
AutoTS-0.3.5-py3-none-any.whl(402.72 KB)
AutoTS-0.3.5.tar.gz(387.80 KB)
0.3.4(Aug 22, 2021)
Latest

improvements to joblib parallelized models (not copying the full df)

additonal parameter checks

made "auto" cpu_count even more conservative

improved 'Score' generation. It should now be more equally weighted across metrics.

fixed potential bug for horizontal ensemble selection if perfect forecasts were delivered

Horizontal ensembles now chosen by combination of multiple metrics and metric_weighting (mae, rmse, spl, contour)

re-weighted fillna probabilities for random choice

addressed a few deprecation warnings

new plot_horizontal() function for AutoTS to quickly visual horizontal ensembles

Probabilistic and HDist ensembles are now deprecated (they can still be run by model_forecast but not by AutoTS class)

new introduce_na parameter which makes series more robust to the last values being NaN in final but never in any validation

Mosaic Ensembles! These can offer major improvements to MAE, but are also less stable than horizontal ensembles.

Source code(tar.gz)
Source code(zip)
AutoTS-0.3.4-py3-none-any.whl(399.19 KB)
AutoTS-0.3.4.tar.gz(384.41 KB)
0.3.3(Aug 5, 2021)
Latest

Fixed horizontal ensembles running in univariate cases (they are explicitly multivariate)

'superfast' transformer list added

test on Mac for the first time, everything seems to work except lightgbm

include first actual unittests (from existing test.py runs)

slight change to random template generation to make sure all models are choosen at least once

cleaned up PredictWitch -> model_forecast() a bit so that users can use it to run single models from parameters directly

added load_live_daily() example data and spruced up production_example.py

tried in vain to make a quiet verbosity option for GluonTS

added create_lagged_regressor

added Greykite model (additional regressors not working yet)

fixed regressors bug in Prophet

added a simple plot method to PredictionObject

fix for deprecation warning in GLS

Source code(tar.gz)
Source code(zip)
AutoTS-0.3.3-py3-none-any.whl(394.20 KB)
AutoTS-0.3.3.tar.gz(379.56 KB)
0.3.2(Jul 1, 2021)
Latest

Table of Contents to Extended Tutorial/Readme.md

Production Example

add weights="mean"/median/min/max

UnivariateRegression

fix check_pickle error for ETS

fix error in Prophet with latest version

VisibleDeprecation warning for hidden_layers random choice in sklearn fixed

prefill_na option added to allow quick filling of NaNs if desired (with zeroes for say, sales forecasting)

made horizontal generalization more stable

fixed bug in VAR where failing on data with negatives

Source code(tar.gz)
Source code(zip)
AutoTS-0.3.2-py3-none-any.whl(384.75 KB)
AutoTS-0.3.2.tar.gz(372.35 KB)
0.3.1(Mar 24, 2021)
Latest

Additional models to GluonTS

GeneralTransformer transformation_params - now handle None or empty dict

cleaning up of the appropriately named 'ModelMonster'

improving MotifSimulation

better error message for all models

enable histgradientboost regressor, left it out before thinking it wouldn't stay experimental this long

import_template now has slightly better method input style

allow ensemble parameter to be a list

NumericTransformer

add .fit_transform method

generally more options and speed improvement

added NumericTransformer to future_regressors, should now coerce if they have different dtypes

Source code(tar.gz)
Source code(zip)
AutoTS-0.3.1-py3-none-any.whl(381.01 KB)
AutoTS-0.3.1.tar.gz(369.20 KB)
0.3.0(Jan 24, 2021)
Latest

breaking change to model templates: transformers structure change

grouping no longer used

parameter generation for transformers allowing more possible combinations

transformer_max_depth parameter

Horizontal Ensembles are now much faster by only running models on the subset of series they apply to

general starting template improved and updated to new transformer format

change many np.random to random

random.choices further necessitates python 3.6 or greater

bug fix in Detrend transformer

bug fix in SeasonalDifference transformer

SPL bug fix when NaN in test set

inverse_transform now fills NaN with zero for upper/lower forecasts

expanded model_list aliases, with dedicated module

bug fix (creating 0,0 order) and tuning of VARMAX

Fix export_template bug

restructuring of some lower-level function locations

Source code(tar.gz)
Source code(zip)
AutoTS-0.3.0-py3-none-any.whl(377.82 KB)
AutoTS-0.3.0.tar.gz(365.89 KB)
0.2.8(Dec 13, 2020)
Latest

Round transformer to replace coerce_integer, ClipOutliers expanded, Slice to replace context_slicer

pd.df Interpolate methods added to FillNA options, " " to "_" in names, rolling_mean_24

slight improvement to printed progress messages

transformer_list (also takes a dict of value:probability) allows adjusting which transformers are created in new generations.

this does not apply to transformers loaded from imported templates

Source code(tar.gz)
Source code(zip)
AutoTS-0.2.8-py3-none-any.whl(372.94 KB)
AutoTS-0.2.8.tar.gz(361.95 KB)
0.2.7(Nov 30, 2020)
Latest

2x speedup in transformation runtime by removing double transformation

joblib parallel to UnobservedComponents

ClipOutliers transformer, Discretize Transformer, CenterLastValue - added in prep for transform template change

bug fix on IntermittentOccurence

minor changes to ETS, now replaces single series failure with zero fill, damped now is damped_trend

0.3.0 is expected to feature a breaking change to model templates in the transformation/pre-processing

Source code(tar.gz)
Source code(zip)
AutoTS-0.2.7-py3-none-any.whl(371.33 KB)
AutoTS-0.2.7.tar.gz(360.35 KB)
0.2.6(Oct 25, 2020)
Latest

fix verbose > 2 error in auto_model

use of f-strings to print some error messages. Python 3.5 may see more complicated error messages as a result.

improved BestN (formery Best3) Ensembles, ensemble collected in dicts

made Horizontal and BestN ensembles tolerant of a component model failure

made Horizontal models capable of generalizing from a subset of series

added info to model table for models that can use future_regressor

added Datepart Regression model, sklearn regressor on time components only

Source code(tar.gz)
Source code(zip)
AutoTS-0.2.6-py3-none-any.whl(369.82 KB)
AutoTS-0.2.6.tar.gz(358.84 KB)
0.2.5(Oct 7, 2020)
Latest

fix error where wide data import skipped cleaning steps

long=True/False for all example data

Source code(tar.gz)
Source code(zip)
AutoTS-0.2.5-py3-none-any.whl(376.46 KB)
AutoTS-0.2.5.tar.gz(972.40 KB)
0.2.4(Sep 30, 2020)
Latest

ARIMA to ARIMAX (for Statsmodels v0.13)

ARIMA parallelization

update of daily sample data, reduced space used by yearly and hourly

n_jobs = 'auto'

models table to extended_tutorial.md

Source code(tar.gz)
Source code(zip)
AutoTS-0.2.4-py3-none-any.whl(376.01 KB)
AutoTS-0.2.4.tar.gz(972.29 KB)
0.2.3(Sep 23, 2020)
Latest

Added n_jobs parameters to pass through to joblib (although a joblib context manager is perhaps the best way)

Added joblib multiprocessing to ETS, GLM, and FBProphet

Fixed future warnings with pandas.DatetimeIndex.week and changes to statsmodels ETS

standardized source code formatting

Source code(tar.gz)
Source code(zip)
AutoTS-0.2.3-py3-none-any.whl(587.24 KB)
AutoTS-0.2.3.tar.gz(1.16 MB)
0.2.2(Jun 28, 2020)
Latest:

grouping/hierarchial reconciliation to GeneralTransformer

allow wide-style data as input

iterative imputer

allow a list of intervals to prediction_intervals in .predict()

Source code(tar.gz)
Source code(zip)
AutoTS-0.2.2-py3-none-any.whl(584.45 KB)
AutoTS-0.2.2.tar.gz(1.15 MB)
0.2.1(Jun 5, 2020)
added 'coerce_integer' to GeneralTransformer

various minor performance improvements

Source code(tar.gz)
Source code(zip)
AutoTS-0.2.1-py3-none-any.whl(581.39 KB)
AutoTS-0.2.1.tar.gz(1.14 MB)

0.2.0(May 31, 2020)

Latest:

Added Github Pages documentation
Changed default for `series_id` so it is no longer required if univariate
Changed default of `subset` to None.
Removed `weighted` parameter, now passing weights to .fit() alone is sufficient.
Fixed a bug where 'One or more series is 90% or more NaN' was printing when it shouldn't
Fixed (or more accurately, reduced) a bug where multiple initial runs were counting as validation runs.
Fixed bug where validation subsetting was behaving oddly
Fixed bug where regressor wasn't being passed to validation.
Renamed preord_ to future_ regressor.
Renamed sample datasets.
Allowed export of result_file as .pickle along with more complete object.
Added model_interrupt parameter to allow for manually skipping models when enabled.
Made serious efforts to make the code prettier with pylint, still lots to do, however...
Improved genetic recombination so optimal models should be reached more quickly
Improved Point to Probabilistic methods:
	'historic_quantile' more stable quantile-based error ranges
	'inferred normal' Bayesian-inspired method
Metrics:
	Added Scaled Pinball Loss (SPL)
	Removed upper/lower MAE
Improved ensembling with new parameter options
	Recursive ensembling (ensemble of ensembles) now enabled
Validation:
	Added 'seasonal' validation method
Categorical transformer improved, now tolerant to leaving bounds.
Added remove_leading_zeroes option for convenience.
Added a number of new Transformer options
	Multiple new Sklearn-sourced transformers (QuantileTransformer, etc)
	SinTrend
	DifferencedDetrend
	CumSumTransformer
	PctChangeTransformer
	PositiveShift Transformer
	Log
	IntermittentOccurrence
	SeasonalDetrend
	bkfilter and cffilter
	DatepartRegression
Entirely changed the general transformer to add ~~three~~ four levels of transformation.
Allowed context_slicer to receive direct integer inputs
Added new 'Detrend' options to allow more sklearn linear models.
GLM
	Error where it apparently won't tolerate any zeroes was compensated for.
	Speed improvement.
RollingRegression
	Added SVM model
	Added option to tune some model parameters to sklearn
	Added new feature construction parameters
	Added RNNs with Keras
GluonTS:
	fixed the use of context_length, added more options to that param
Dynamic Factor added uncertainty from Statsmodels Statespace
VARMAX added uncertainty from Statsmodels Statespace
New models:
	SeasonalNaive model
	VAR from Statsmodels (faster than VARMAX statespace)
	MotifSimulation
	WindowRegression
	TensorflowSTS
	TFPRegression
	ComponentAnalysis

Source code(tar.gz)
Source code(zip)
AutoTS-0.2.0-py3-none-any.whl(580.53 KB)
AutoTS-0.2.0.tar.gz(1.14 MB)

0.2.0a2(May 22, 2020)

Alpha release of major version release 0.2.0
Source code(tar.gz)
Source code(zip)
AutoTS-0.2.0a2.tar.gz(1.13 MB)

Automated Time Series Forecasting

Related tags

Overview

AutoTS

Table of Contents

Installation

Basic Use

Tips for Speed and Large Data:

How to Contribute:

Comments

Detected Unintuitive Behaviour

Root Cause

Proposed Change

Name Version Build Channel

Releases(0.5.3)

0.5.3(Dec 23, 2022)

0.5.3 🌍 🌍 🌍

0.5.2(Dec 13, 2022)

0.5.2 🌎 🌎 🌎

0.5.1(Nov 15, 2022)

0.5.1 :mage_man: :mage_man: :mage_man:

0.5.0(Aug 24, 2022)

0.5.0 :space_invader: :space_invader: :space_invader:

0.4.2(Jun 20, 2022)

Latest :space_invader: :space_invader: :space_invader:

0.4.1(May 16, 2022)

Latest :fireworks: :fireworks:

0.4.0(Feb 7, 2022)

Latest

0.3.12(Dec 5, 2021)

Latest

0.3.11(Nov 29, 2021)

Latest

0.3.10(Nov 4, 2021)

Latest

0.3.9(Oct 31, 2021)

Latest

0.3.8(Oct 17, 2021)

Latest

0.3.7(Oct 4, 2021)

Latest

0.3.6(Sep 14, 2021)

Latest

0.3.5(Aug 30, 2021)

Latest

0.3.4(Aug 22, 2021)

Latest

0.3.3(Aug 5, 2021)

Latest

0.3.2(Jul 1, 2021)

Latest

0.3.1(Mar 24, 2021)

Latest

0.3.0(Jan 24, 2021)

Latest

0.2.8(Dec 13, 2020)

Latest

0.2.7(Nov 30, 2020)

Latest

0.2.6(Oct 25, 2020)

Latest

0.2.5(Oct 7, 2020)

Latest

0.2.4(Sep 30, 2020)

Latest

0.2.3(Sep 23, 2020)

Latest

0.2.2(Jun 28, 2020)

Latest:

0.2.1(Jun 5, 2020)

0.2.0(May 31, 2020)

0.2.0a2(May 22, 2020)

Owner

Colin Catlin

Time series forecasting with PyTorch

ETNA is an easy-to-use time series forecasting framework.

Nixtla is an open-source time series forecasting library.

ETNA – time series forecasting framework

Continuously evaluated, functional, incremental, time-series forecasting

AtsPy: Automated Time Series Models in Python (by @firmai)