Time Series Feature Extraction

Computes per-series summary features from the target variable via tsfeatures::tsfeatures() and broadcasts them as constant columns to every row of the corresponding series. For an unkeyed task the features are broadcast to every row; for a keyed task each key contributes one feature vector.

Features are cached in the state at train time and reused at predict time. Predicting on a key that was not seen during training is an error.

Parameters

The parameters are the parameters inherited from mlr3pipelines::PipeOpTaskPreproc, as well as:

• features :: character()
  Function names from the tsfeatures namespace that return numeric feature vectors. Default c("frequency", "stl_features", "entropy", "acf_features").

• scale :: logical(1)
  If TRUE, scale each series to mean 0 and sd 1 before feature extraction. Default TRUE.

• trim :: logical(1)
  If TRUE, trim values outside ±trim_amount before feature extraction. Default FALSE.

• trim_amount :: numeric(1)
  Trimming threshold. Default 0.1.

• parallel :: logical(1)
  If TRUE, compute features in parallel via a future::plan(). Default FALSE.

• multiprocess :: function
  Function from the future package used when parallel = TRUE. Default future::multisession().

• na.action :: function
  Missing-value handler. Default stats::na.pass().

Super classes

mlr3pipelines::PipeOp -> mlr3pipelines::PipeOpTaskPreproc -> PipeOpFcstTsfeats

Methods

Public methods

• PipeOpFcstTsfeats$new()

• PipeOpFcstTsfeats$clone()

Inherited methods

• mlr3pipelines::PipeOp$help()
• mlr3pipelines::PipeOp$predict()
• mlr3pipelines::PipeOp$print()
• mlr3pipelines::PipeOp$train()

---

Method new()

Initializes a new instance of this Class.

Usage

PipeOpFcstTsfeats$new(id = "fcst.tsfeats", param_vals = list())

Arguments

id

(character(1))
Identifier of resulting object, default "fcst.tsfeats".

param_vals

(named list())
List of hyperparameter settings, overwriting the hyperparameter settings that would otherwise be set during construction. Default list().

---

Method clone()

The objects of this class are cloneable with this method.

Usage

PipeOpFcstTsfeats$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

if (FALSE) { # \dontrun{
library(mlr3pipelines)
task = tsk("airpassengers")
po = po("fcst.tsfeats", features = c("entropy", "acf_features"))
out = po$train(list(task))[[1L]]
out$head()
} # }