Time series transformations#
The aeon.transformations
module contains classes for data
transformations.
Transformations.
All (simple) transformers in aeon``can be listed using the ``aeon.registry.all_estimators
utility,
using estimator_types="regressor"
, optionally filtered by tags.
Valid tags can be listed using aeon.registry.all_tags
.
For pairwise transformers (time series distances, kernels), instead see _transformations_pairwise_ref.
Transformations are categorized as follows:
Category |
Explanation |
Example |
---|---|---|
Composition |
Building blocks for pipelines, wrappers, adapters |
Transformer pipeline |
Series-to-features |
Transforms series to float/category vector |
Length and mean |
Series-to-series |
Transforms individual series to series |
Differencing, detrending |
Series-to-Panel |
transforms a series into a panel |
Bootstrap, sliding window |
Panel transform |
Transforms panel to panel, not by-series |
Padding to equal length |
Hierarchical |
uses hierarchy information non-trivially |
Reconciliation |
Composition#
Pipeline building#
|
Pipeline of transformers compositor. |
|
Concatenates results of multiple transformer objects. |
|
Apply a transformer columnwise to multivariate series. |
|
Transformer wrapper to delay fit to the transform phase. |
|
Facilitate an AutoML based selection of the best transformer. |
|
Wrap an existing transformer to tune whether to include it in a pipeline. |
|
Invert a series-to-series transformation. |
|
Identity transformer, returns data unchanged in transform/inverse_transform. |
|
Create exogeneous features which are a copy of the endogenous data. |
|
Constructs a transformer from an arbitrary callable. |
Sklearn and pandas adapters#
A transformer that turns time series collection into tabular data. |
|
|
Turns time series collections data into tabular data based on intervals. |
|
Adapt scikit-learn transformation interface to time series setting. |
|
Adapt pandas transformations to aeon interface. |
Series-to-features transformers#
Series-to-features transformers transform individual time series to a collection of primitive features. Primitive features are usually a vector of floats, but can also be categorical.
When applied to panels or hierarchical data, the transformation result is a table with as many rows as time series in the collection.
Summarization#
These transformers extract simple summary features.
|
Calculate summary value of a time series. |
|
Transformer for extracting time series features. |
|
Fitted parameter extractor. |
Shapelets, wavelets, and convolution#
|
Random Shapelet Transform. |
Random Dilated Shapelet Transform (RDST) as described in [R1a26faa97573-1]_[R1a26faa97573-2]_. |
|
RandOm Convolutional KErnel Transform (ROCKET). |
|
MINImally RandOm Convolutional KErnel Transform (MiniRocket). |
|
MINImally RandOm Convolutional KErnel Transform (MiniRocket) multivariate. |
MINIROCKET (Multivariate, unequal length). |
|
|
Multi RandOm Convolutional KErnel Transform (MultiRocket). |
|
Multi RandOm Convolutional KErnel Transform (MultiRocket). |
|
Discrete Wavelet Transform Transformer. |
Distance-based features#
|
Return the matrix profile and index profile for each time series of a dataset. |
Dictionary-based features#
Moment-based features#
|
Transformation class from the signature method. |
Feature collections#
These transformers extract larger collections of features.
Transformer for extracting time series features via tsfresh.extract_features. |
|
|
Transformer for extracting time series features via tsfresh.extract_features. |
|
Canonical Time-series Characteristics (Catch22). |
Series-to-series transformers#
Series-to-series transformers transform individual time series into another time series.
When applied to panels or hierarchical data, individual series are transformed.
Lagging#
|
Lagging transformer. |
Element-wise transforms#
These transformations apply a function element-wise.
Depending on the transformer, the transformation parameters can be fitted.
|
Box-Cox power transform. |
|
Natural logarithm transformation. |
|
Scaled logit transform or Log transform. |
|
Cosine transformation. |
|
Apply element-wise exponentiation transformation to a time series. |
|
Apply element-sise square root transformation to a time series. |
Detrending#
|
Remove a trend from a series. |
|
Remove seasonal components from a time series. |
|
Remove seasonal components from time series, conditional on seasonality test. |
|
Remove seasonal components from a time-series using STL. |
|
Clear sky transformer for solar data. |
Filtering and denoising#
|
Filter a times series using the Baxter-King filter. |
|
Transformer that filters Series data. |
|
Kalman Filter Transformer. |
|
Decompose the original data into two or more Theta-lines. |
Differencing and slope#
|
Apply iterative differences to a timeseries. |
|
Piecewise slope transformation. |
Binning and segmentation#
|
Bins time series and aggregates by bin. |
|
Time series interpolator/re-sampler. |
|
Interval segmentation transformer. |
|
Random interval segmenter transformer. |
Missing value imputation#
|
Missing value imputation. |
Seasonality and Date-Time Features#
|
DateTime feature extraction for use in e.g. tree based models. |
|
Get Element-wise time elapsed between the timeindex and a reference start time. |
|
Fourier Features for time series seasonality. |
Auto-correlation series#
|
Auto-correlation transformer. |
|
Partial auto-correlation transformer. |
Window-based series transforms#
These transformers create a series based on a sequence of sliding windows.
|
Calculate the matrix profile of a time series. |
|
HOG1D transform. |
Multivariate-to-univariate#
These transformers convert multivariate series to univariate.
|
Concatenate multivariate series to a long univariate series. |
Augmentation#
Augmenter inverting the time series by multiplying it by -1. |
|
|
Draw random samples from time series. |
Augmenter reversing the time series. |
|
|
Augmenter adding Gaussian (i.e. white) noise to the time series. |
FeatureSelection#
These transformers select features in X based on y.
|
Select exogenous features. |
|
Elbow Class Sum (ECS) transformer to select a subset of channels/variables. |
|
Elbow Class Pairwise (ECP) transformer to select a subset of channels. |
Subsetting time points and variables#
These transformers subset X by time points (pandas index or index level) or variables (pandas columns).
|
Column selection transformer. |
|
Index subsetting transformer. |
Panel transformers#
Panel transformers transform a panel of time series into a panel of time series.
A panel transformer is fitted on an entire panel, and not per series.
Equal length transforms#
These transformations ensure all series in a panel have equal length
|
Pad unequal length time series to equal, fixed length. |
|
Truncate unequal length time series to a lower bounds. |
Dimension reduction#
|
Principal Components Analysis applied as transformer. |
Series-to-Panel transformers#
These transformers create a panel from a single series.
Bootstrap transformations#
|
Creates a population of similar time series. |
|
Moving Block Bootstrapping method for synthetic time series generation. |
Outlier detection, changepoint detection#
|
Use HampelFilter to detect outliers based on a sliding window. |
|
ClaSP (Classification Score Profile) Transformer. |
Hierarchical transformers#
These transformers are specifically for hierarchical data and panel data.
The transformation depends on the specified hierarchy in a non-trivial way.
|
Prepare hierarchical data, including aggregate levels, from bottom level. |
|
Hierarchical reconcilation transformer. |