crossv_tskfold: Generate cross-validated time-series K-fold test-training...

In jrnold/resamplr: Resampling methods

Description

Splits data into test/train sets for cross-validating time series data. In each split, the indices are increasing, so only previous data are used to test future data.

Usage

crossv_tskfold(data, K, ...)

## S3 method for class 'data.frame'
crossv_tskfold(data, K = 5L, ...)

## S3 method for class 'grouped_df'
crossv_tskfold(data, K = 5L, ...)

Arguments

data	A data frame or vector
K	The number of folds
...	Passed to methods

Details

The data are split into K ≥q 2 folds, where for each i in 2:K, the i-th fold is the test set and folds 1, ..., (i - 1) folds are the training set.

Value

A data frame with K the following columns:

sample

A list of resample objects. Training sets.

.id

An integer vector of identifiers.

Methods (by class)

• data.frame: Rows (observations) are split into folds. The ordering of the data frame is assmed to be meaningful.

• grouped_df: The groups are split into folds. The ordering of the groups is assumed to be meaningful. No stratification option is provided since the cases in which stratification can usually be rewritten to operate on the groups rather than within groups.

References

• Scikit-learn v. 0.18.1. Cross-validation: evaluating estimator performance. URL.

jrnold/resamplr documentation built on May 20, 2019, 1:05 a.m.