API Reference

This is the class and function reference for pmdarima. Please refer to the full user guide for further details, as the class and function raw specifications may not be enough to give full guidelines on their uses.

`pmdarima.arima`: ARIMA estimator & differencing tests

The pmdarima.arima sub-module defines the ARIMA estimator and the auto_arima function, as well as a set of tests of seasonality and stationarity.

ARIMA estimator & statistical tests

User guide: See the Estimating the seasonal differencing term, D and Enforcing stationarity sections for further details.

`arima.ADFTest`([alpha, k])	Conduct an ADF test for stationarity.
`arima.ARIMA`(order[, seasonal_order, ...])	An ARIMA estimator.
`arima.AutoARIMA`([start_p, d, start_q, ...])	Automatically discover the optimal order for an ARIMA model.
`arima.CHTest`(m)	Conduct a CH test for seasonality.
`arima.KPSSTest`([alpha, null, lshort])	Conduct a KPSS test for stationarity.
`arima.OCSBTest`(m[, lag_method, max_lag])	Perform an OCSB test of seasonality.
`arima.PPTest`([alpha, lshort])	Conduct a PP test for stationarity.
`arima.StepwiseContext`([max_steps, max_dur])	Context manager to capture runtime context for stepwise mode.

ARIMA auto-parameter selection

User guide: See the Tips to using auto_arima section for further details.

arima.auto_arima(y[, X, start_p, d, ...])

Automatically discover the optimal order for an ARIMA model.

Differencing helpers

`arima.is_constant`(x)	Test `x` for constancy.
`arima.ndiffs`(x[, alpha, test, max_d])	Estimate ARIMA differencing term, `d`.
`arima.nsdiffs`(x, m[, max_D, test])	Estimate the seasonal differencing term, `D`.

Seasonal decomposition

arima.decompose(x, type_, m[, filter_])

Decompose the time series into trend, seasonal, and random components.

`pmdarima.datasets`: Toy timeseries datasets

The pmdarima.datasets submodule provides several different univariate time- series datasets used in various examples and tests across the package. If you would like to prototype a model, this is a good place to find easy-to-access data.

User guide: See the Toy time-series datasets section for further details.

Dataset loading functions

`datasets.load_airpassengers`([as_series, dtype])	Monthly airline passengers.
`datasets.load_ausbeer`([as_series, dtype])	Quarterly beer production data.
`datasets.load_austres`([as_series, dtype])	Quarterly residential data.
`datasets.load_gasoline`([as_series, dtype])	Weekly US finished motor gasoline products
`datasets.load_heartrate`([as_series, dtype])	Uniform heart-rate data.
`datasets.load_lynx`([as_series, dtype])	Annual numbers of lynx trappings for 1821–1934 in Canada.
`datasets.load_msft`()	Load the microsoft stock data
`datasets.load_sunspots`([as_series, dtype])	Monthly Sunspot Numbers, 1749 - 1983
`datasets.load_taylor`([as_series, dtype])	Half-hourly electricity demand
`datasets.load_wineind`([as_series, dtype])	Australian total wine sales by wine makers in bottles <= 1 litre.
`datasets.load_woolyrnq`([as_series, dtype])	Quarterly production of woollen yarn in Australia.

`pmdarima.metrics`: Time-series metrics

The metrics submodule implements time-series metrics that are not implemented in scikit-learn.

Metrics

metrics.smape(y_true, y_pred)

Compute the Symmetric Mean Absolute Percentage Error.

`pmdarima.model_selection`: Cross-validation classes

The pmdarima.model_selection submodule defines several different strategies for cross-validating time series models

Cross validation & split utilities

`model_selection.RollingForecastCV`([h, step, ...])	Use a rolling forecast to perform cross validation
`model_selection.SlidingWindowForecastCV`([h, ...])	Use a sliding window to perform cross validation

`model_selection.check_cv`([cv])	Input checker utility for building a cross-validator
`model_selection.cross_validate`(estimator, y)	Evaluate metric(s) by cross-validation and also record fit/score times.
`model_selection.cross_val_predict`(estimator, y)	Generate cross-validated estimates for each input data point
`model_selection.cross_val_score`(estimator, y)	Evaluate a score by cross-validation
`model_selection.train_test_split`(*arrays[, ...])	Split arrays or matrices into sequential train and test subsets

`pmdarima.pipeline`: Pipelining transformers & ARIMAs

With the pipeline.Pipeline class, we can pipeline transformers together and into a final ARIMA stage.

Pipelines

pipeline.Pipeline(steps)

A pipeline of transformers with an optional final estimator stage

`pmdarima.preprocessing`: Preprocessing transformers

The pmdarima.preprocessing submodule provides a number of transformer classes for pre-processing time series or exogenous arrays.

Endogenous transformers

`preprocessing.BoxCoxEndogTransformer`([...])	Apply the Box-Cox transformation to an endogenous array
`preprocessing.LogEndogTransformer`([lmbda, ...])	Apply a log transformation to an endogenous array

Exogenous transformers

`preprocessing.DateFeaturizer`(column_name[, ...])	Create exogenous date features
`preprocessing.FourierFeaturizer`(m[, k, prefix])	Fourier terms for modeling seasonality

`pmdarima.utils`: Utilities

Utilities and array differencing functions used commonly across the package.

Array helper functions & metaestimators

`utils.acf`(x[, nlags, qstat, fft, alpha, ...])	Calculate the autocorrelation function.
`utils.as_series`(x, **kwargs)	Cast as pandas Series.
`utils.c`(*args)	Imitates the `c` function from R.
`utils.check_endog`(y[, dtype, copy, ...])	Wrapper for `check_array` and `column_or_1d` from sklearn
`utils.diff`(x[, lag, differences])	Difference an array.
`utils.diff_inv`(x[, lag, differences, xi])	Inverse the difference of an array.
`utils.if_has_delegate`(delegate)	Wrap a delegated instance attribute function.
`utils.is_iterable`(x)	Test a variable for iterability.
`utils.pacf`(x[, nlags, method, alpha])	Partial autocorrelation estimate.

Plotting utilities & wrappers

`utils.autocorr_plot`(series[, show])	Plot a series' auto-correlation.
`utils.decomposed_plot`(decomposed_tuple[, ...])	Plot the decomposition of a time series.
`utils.plot_acf`(series[, ax, lags, alpha, ...])	Plot a series' auto-correlation as a line plot.
`utils.plot_pacf`(series[, ax, lags, alpha, ...])	Plot a series' partial auto-correlation as a line plot.