demo/forest_demo.R

In setartree: SETAR-Tree - A Novel and Accurate Tree Algorithm for Global Time Series Forecasting

library(setartree)


### Case 1 - When need to obtain forecasts for a list of time series ###

# Training SETAR-Forest
forest1 <- setarforest(chaotic_logistic_series, bagging_freq = 3)

# Obtaining predictions
forest_forecasts <- forecast(forest1, chaotic_logistic_series)

# Plotting - can use either plot or autoplot functions in the 'forecast' package
tryCatch({
  if(!is.null(find.package("forecast"))){
    library(forecast) # The 'forecast' package is only required for plotting examples shown in lines 17-30.

    plot(forest_forecasts) # Plot all time series
    autoplot(forest_forecasts)

    plot(forest_forecasts$forecast$T1) # Plot the 1st time series
    autoplot(forest_forecasts$forecast$T1)

    plot(forest_forecasts$forecast$T10) # Plot the 10th time series
    autoplot(forest_forecasts$forecast$T10)

    custom_forecasts <- forest_forecasts # Plot a selected set of time series - eg: 1st and 10th time series
    custom_forecasts$forecast <- custom_forecasts$forecast[c(1, 10)]
    custom_forecasts$method <- custom_forecasts$method[c(1, 10)]
    plot(custom_forecasts)
    autoplot(custom_forecasts)
  }
},error = function(e) {})



### Case 2 - When need to obtain predictions for a dataframe/matrix containing model inputs including past time series lags and numerical/categorical covariates ###

# Training SETAR-Forest
forest2 <- setarforest(data = web_traffic_train[-1],
                       label = web_traffic_train[,1],
                       bagging_freq = 3,
                       stopping_criteria = "both",
                       categorical_covariates = "Project")

# Obtaining predictions
forecast(forest2, web_traffic_test)