window_reg() is a way to generate a specification of a window model
before fitting and allows the model to be created using
different backends.
Arguments
- mode
A single character string for the type of model. The only possible value for this model is "regression".
- id
An optional quoted column name (e.g. "id") for identifying multiple time series (i.e. panel data).
- window_size
A window to apply the window function. By default, the window uses the full data set, which is rarely the best choice.
Details
A time series window regression is derived using window_reg().
The model can be created using the fit() function using the
following engines:
"window_function" (default) - Performs a Window Forecast applying a
window_function(engine parameter) to a window of size defined bywindow_size
Engine Details
function (default engine)
The engine uses window_function_fit_impl(). A time series window function
applies a window_function to a window of the data (last N observations).
The function can return a scalar (single value) or multiple values that are repeated for each window
Common use cases:
Moving Average Forecasts: Forecast forward a 20-day average
Weighted Average Forecasts: Exponentially weighting the most recent observations
Median Forecasts: Forecasting forward a 20-day median
Repeating Forecasts: Simulating a Seasonal Naive Forecast by broadcasting the last 12 observations of a monthly dataset into the future
The key engine parameter is the window_function. A function / formula:
If a function, e.g.
mean, the function is used with any additional arguments,...inset_engine().If a formula, e.g.
~ mean(., na.rm = TRUE), it is converted to a function.
This syntax allows you to create very compact anonymous functions.
Fit Details
Date and Date-Time Variable
It's a requirement to have a date or date-time variable as a predictor.
The fit() interface accepts date and date-time features and handles them internally.
fit(y ~ date)
ID features (Multiple Time Series, Panel Data)
The id parameter is populated using the fit() or fit_xy() function:
ID Example: Suppose you have 3 features:
y(target)date(time stamp),series_id(a unique identifer that identifies each time series in your data).
The series_id can be passed to the window_reg() using
fit():
window_reg(id = "series_id")specifes that theseries_idcolumn should be used to identify each time series.fit(y ~ date + series_id)will passseries_idon to the underlying functions.
Window Function Specification (window_function)
You can specify a function / formula using purrr syntax.
If a function, e.g.
mean, the function is used with any additional arguments,...inset_engine().If a formula, e.g.
~ mean(., na.rm = TRUE), it is converted to a function.
This syntax allows you to create very compact anonymous functions.
Window Size Specification (window_size)
The period can be non-seasonal (window_size = 1 or "none") or
yearly seasonal (e.g. For monthly time stamps, window_size = 12, window_size = "12 months", or window_size = "yearly").
There are 3 ways to specify:
window_size = "all": A seasonal period is selected based on the periodicity of the data (e.g. 12 if monthly)window_size = 12: A numeric frequency. For example, 12 is common for monthly datawindow_size = "1 year": A time-based phrase. For example, "1 year" would convert to 12 for monthly data.
External Regressors (Xregs)
These models are univariate. No xregs are used in the modeling process.
Examples
library(dplyr)
library(parsnip)
library(rsample)
library(timetk)
# Data
m750 <- m4_monthly %>% filter(id == "M750")
m750
#> # A tibble: 306 × 3
#> id date value
#> <fct> <date> <dbl>
#> 1 M750 1990-01-01 6370
#> 2 M750 1990-02-01 6430
#> 3 M750 1990-03-01 6520
#> 4 M750 1990-04-01 6580
#> 5 M750 1990-05-01 6620
#> 6 M750 1990-06-01 6690
#> 7 M750 1990-07-01 6000
#> 8 M750 1990-08-01 5450
#> 9 M750 1990-09-01 6480
#> 10 M750 1990-10-01 6820
#> # ℹ 296 more rows
# Split Data 80/20
splits <- initial_time_split(m750, prop = 0.8)
# ---- WINDOW FUNCTION -----
# Used to make:
# - Mean/Median forecasts
# - Simple repeating forecasts
# Median Forecast ----
# Model Spec
model_spec <- window_reg(
window_size = 12
) %>%
# Extra parameters passed as: set_engine(...)
set_engine(
engine = "window_function",
window_function = median,
na.rm = TRUE
)
# Fit Spec
model_fit <- model_spec %>%
fit(log(value) ~ date, data = training(splits))
model_fit
#> parsnip model object
#>
#> WINDOW FUNC [12]
#> --------
#> Model:
#> # A tibble: 1 × 1
#> value
#> <dbl>
#> 1 9.26
# Predict
# - The 12-month median repeats going forward
predict(model_fit, testing(splits))
#> # A tibble: 62 × 1
#> .pred
#> <dbl>
#> 1 9.26
#> 2 9.26
#> 3 9.26
#> 4 9.26
#> 5 9.26
#> 6 9.26
#> 7 9.26
#> 8 9.26
#> 9 9.26
#> 10 9.26
#> # ℹ 52 more rows
# ---- PANEL FORECAST - WINDOW FUNCTION ----
# Weighted Average Forecast
model_spec <- window_reg(
# Specify the ID column for Panel Data
id = "id",
window_size = 12
) %>%
set_engine(
engine = "window_function",
# Create a Weighted Average
window_function = ~ sum(tail(.x, 3) * c(0.1, 0.3, 0.6)),
)
# Fit Spec
model_fit <- model_spec %>%
fit(log(value) ~ date + id, data = training(splits))
model_fit
#> parsnip model object
#>
#> WINDOW FUNC [12]
#> --------
#> Model:
#> # A tibble: 1 × 2
#> id value
#> <fct> <dbl>
#> 1 M750 9.29
# Predict: The weighted average (scalar) repeats going forward
predict(model_fit, testing(splits))
#> # A tibble: 62 × 1
#> .pred
#> <dbl>
#> 1 9.29
#> 2 9.29
#> 3 9.29
#> 4 9.29
#> 5 9.29
#> 6 9.29
#> 7 9.29
#> 8 9.29
#> 9 9.29
#> 10 9.29
#> # ℹ 52 more rows
# ---- BROADCASTING PANELS (REPEATING) ----
# Simulating a Seasonal Naive Forecast by
# broadcasted model the last 12 observations into the future
model_spec <- window_reg(
id = "id",
window_size = Inf
) %>%
set_engine(
engine = "window_function",
window_function = ~ tail(.x, 12),
)
# Fit Spec
model_fit <- model_spec %>%
fit(log(value) ~ date + id, data = training(splits))
model_fit
#> parsnip model object
#>
#> WINDOW FUNC [Inf]
#> --------
#> Model:
#> # A tibble: 12 × 2
#> id value
#> <fct> <dbl>
#> 1 M750 9.27
#> 2 M750 9.27
#> 3 M750 9.15
#> 4 M750 9.19
#> 5 M750 9.18
#> 6 M750 9.25
#> 7 M750 9.26
#> 8 M750 9.27
#> 9 M750 9.26
#> 10 M750 9.26
#> 11 M750 9.29
#> 12 M750 9.29
# Predict: The sequence is broadcasted (repeated) during prediction
predict(model_fit, testing(splits))
#> # A tibble: 62 × 1
#> .pred
#> <dbl>
#> 1 9.27
#> 2 9.27
#> 3 9.15
#> 4 9.19
#> 5 9.18
#> 6 9.25
#> 7 9.26
#> 8 9.27
#> 9 9.26
#> 10 9.26
#> # ℹ 52 more rows