Fit a random forest model with the output as the 1-step ahead
value and the predictors as the 5 previous lags. This is functionally
similar to the time-delay embedding approach with an embedding of 5, but
using randomForest to estimate the forecast
function.
ranfor_ts(timeseries, num_ahead = 5, level = 95)
| timeseries | the time series to forecast |
|---|---|
| num_ahead | the number of points at the end of the time series to forecast |
| level | Confidence level for prediction intervals. |
a data.frame of the mean forecasts, the observed values, and the lower and upper CI levels (if an error occurs, then just NA values)