Time Series Forecasting Using 23 Individual Models.

ForecastingEnsembles

The goal of ForecastingEnsembles is to automatically run 23 forecasting models, return 19 plots and tables.

Installation

You can install the development version of ForecastingEnsembles like so:

devtools::install_github("InfiniteCuriosity/ForecastingEnsembles")

Example

This is a basic example which shows you how to solve a common problem of forecasting data that is reported weekly, monthly or quarterly. The Forecasting Ensembles package only needs the data, and it will build 23 models:

Linear1 = TSLM(Value ~ season() + trend()),
Linear2 = TSLM(Value),
Linear3 = TSLM(Value ~ season()),
Linear4 = TSLM(Value ~ trend()),
Arima1 = ARIMA(Value ~ season() + trend(), stepwise = TRUE, greedy = TRUE, approximation = TRUE),
Arima2 = ARIMA(Value ~ season(), stepwise = TRUE, greedy = TRUE, approximation = TRUE),
Arima3 = ARIMA(Value ~ trend(), stepwise = TRUE, greedy = TRUE, approximation = TRUE),
Arima4 = ARIMA(Value),
Deterministic = ARIMA(Value ~ 1 + pdq(d = 0), stepwise = TRUE, greedy = TRUE, approximation = TRUE),
Stochastic = ARIMA(Value ~ pdq(d = 1), stepwise = TRUE, greedy = TRUE, approximation = TRUE),
Ets1 = ETS(Value ~ season() + trend()),
Ets2 = ETS(Value ~ trend()),
Ets3 = ETS(Value ~ season()),
Ets4 = ETS(Value),
Holt_Winters_Additive = ETS(Value ~ error("A") + trend("A") + season("A")),
Holt_Winters_Multiplicative = ETS(Value ~ error("M") + trend("A") + season("M")),
Holt_Winters_Damped = ETS(Value ~ error("M") + trend("Ad") + season("M")),
Fourier1 = ARIMA((Value) ~ fourier(K = 1) + PDQ(0, 0, 0)),
Fourier2 = ARIMA((Value) ~ fourier(K = 2) + PDQ(0, 0, 0)),
Fourier3 = ARIMA((Value) ~ fourier(K = 3) + PDQ(0, 0, 0)),
Fourier4 = ARIMA((Value) ~ fourier(K = 4) + PDQ(0, 0, 0)),
Fourier5 = ARIMA((Value) ~ fourier(K = 5) + PDQ(0, 0, 0)),
Fourier6 = ARIMA((Value) ~ fourier(K = 6) + PDQ(0, 0, 0))

library(ForecastingEnsembles)
## Forecasting(time_series_data = Oct_2024_all_nonfarm, train_amount = 0.60, number_of_intervals_to_forecast = 1, use_parallel = "Y", time_interval = "M")

The Forecasting Ensembles package returns:

Plot of the value of the time series Plot of individual seasons Head of the data Tail of the data Plot of the trend Plot of the seasonally adjusted data Plot of the decomposition Plot of anomalies Plot of subseasons Plot of multiple lags Table of time series features Table of time series quartiles Table of the best model forecast Plot of the best model forecast Plot of best autocorrelation function (ACF) Plot of the best model histogram of the residuals Plot of the best model actual vs predicted Plot of the best model actual vs residuals Plot of the best model baseline data.