multivariate_M4benchmarks: multivariate_M4benchmarks

In jdestefani/MM4Benchmark: Multivariate (multi-univariate) M4 Benchmarks

Description

Function producing the multivariate forecasts employing all the benchmark methods of the M4 competition. The multivariate implementation is made by launching in parallel the different univariate M4 benchmarks https://github.com/M4Competition/M4-methods/blob/master/Benchmarks%20and%20Evaluation.R on all the multivariate series

Usage

multivariate_M4benchmarks(Xtr, h)

Arguments

Xtr	Training data (numeric matrix with N columns)
h	Forecasting horizon

Value

List containing:

• Naive: h-step forecast for the naive method (numeric matrix - dimensions h x N)

• NaiveSeasonal: h-step forecast for the seasonal naive method (numeric matrix - dimensions h x N)

• SimpleES: h-step forecast for the simple exponential smoothing method (numeric matrix - dimensions h x N)

• HoltWinters: h-step forecast for the Holt-Winters method (numeric matrix - dimensions h x N)

• HoltWintersDamped: h-step forecast for Holt-Winters damped method (numeric matrix - dimensions h x N)

• Theta: h-step forecast for the theta method (numeric matrix - dimensions h x N)

• Combined: h-step forecast for the combined method (numeric matrix - dimensions h x N)

• TimeNaive: Computational time for the naive method (numeric scalar)

• TimeNaiveSeasonal: Computational time for the seasonal naive method (numeric scalar)

• TimeSimpleES: Computational time for the simple exponential smoothing method (numeric scalar)

• TimeHoltWinters: Computational time for the Holt-Winters method (numeric scalar)

• TimeHoltWintersDamped: Computational time for Holt-Winters damped method (numeric scalar)

• TimeTheta: Computational time for the theta method (numeric scalar)

• TimeCombined: Computational time for the combined method (numeric scalar)

Examples

X <- EuStockMarkets
splitting_point <- round(2*dim(X)[1]/3)
X_train <- X[1:splitting_point,]
h <- 5
multivariate_results <- multivariate_M4benchmarks(X_train,h)

jdestefani/MM4Benchmark documentation built on Dec. 20, 2021, 10:05 p.m.