ts_normalization: Normalize univariate timeseries

In thfuchs/fcf: Time Series Forecasting using Recurrent Neural Networks by Keras

Description

Normalize univariate timeseries

Usage

ts_normalization(
  data,
  length_val,
  length_test,
  value_col = "value",
  joined = TRUE,
  metrics = FALSE
)

Arguments

data	univariate time series (data.frame / data.table)
length_val	length for validation set
length_test	length for test set
value_col	column(s) to normalize, searched by starting pattern. E.g. value_col = "index" will catch column "index" and "index_2" but not "2_index"
joined	joined normalization for same pattern? TRUE by default. See section "Joined value columns" for details
metrics	return data only or list of data and metrics?

Value

Depending on metrics, processed DT object or list of "data" and "metrics" (center and scale)

Joined value columns

Joined means to normalize all columns detected by pattern with the one column exactly matching. Watch out for this condition to hold if joined = TRUE.
joined is of particular use for lagged time series. E.g. column "value" should be used to normalize not only column "value" but also "value_lag1" etc.

Examples

# without metrics
DT_norm <- ts_normalization(tsRNN::DT_apple, 10, 10); DT_norm

# with metrics
ts_normalization(tsRNN::DT_apple, 10, 10, metrics = TRUE)

thfuchs/fcf documentation built on April 18, 2021, 1:43 p.m.