Using regression trees for univariate time series forecasting with the utsf package

This paper presents the R package utsf for univariate time series forecasting based on machine learning regression models. The package offers a simple, unified interface for applying different models in an autoregressive manner with recursive forecasts. Currently, utsf is focused on regression tree models, but its extensible design allows users to apply their own regression models. In doing so, they can take advantage of all the features implemented in the package, such as constructing training sets with autoregressive examples, implementing recursive forecasts, prediction intervals, preprocessing data, estimating forecast accuracy, and tuning parameters.