Forecast future values for NMF-VAR model

nmfkc.ar.predict computes multi-step-ahead forecasts for a fitted NMF-VAR model using recursive forecasting.

If the fitted model contains time series property information (from nmfkc.ar), the forecasted values will have appropriate time-based column names.

Usage

nmfkc.ar.predict(x, Y, degree = NULL, n.ahead = 1)

Arguments

x

An object of class nmfkc (the fitted model).

Y

The historical observation matrix used for fitting (or at least the last degree columns).

degree

Optional integer. Lag order (D). If NULL (default), it is inferred from x$A.attr (when available) or from the dimensions of x$C.

n.ahead

Integer (>=1). Number of steps ahead to forecast.

Value

A list with components:

pred

A \(P \times n.ahead\) matrix of predicted values. Column names are future time points if time information is available.

time

A numeric vector of future time points corresponding to the columns of pred.

See also

nmfkc, nmfkc.ar

Examples

# Forecast AirPassengers
d <- AirPassengers
ar_data <- nmfkc.ar(d, degree = 2)
result <- nmfkc(ar_data$Y, ar_data$A, rank = 1)
#> Y(1,142)~X(1,1)C(1,3)A(3,142)=XB(1,142)...
#> 0sec
pred <- nmfkc.ar.predict(result, Y = matrix(d, nrow = 1), degree = 2, n.ahead = 3)
pred$pred
#>          1961 1961.08333 1961.16667
#> [1,] 427.6732   424.8679   422.1085