We have encountered a set of models for the different regression models thus far. Time series data brings additional complexity, and hence we have even more models to choose from (or rather, ensemble from). A quick review of the important models is provided here. Most of the models discussed here deal with univariate time series , and we need even more specialized models and methods to incorporate . We will begin with the simplest possible time series model and then move up to the neural network implementations.
Suppose that we have the data , and we need forecasts for the next h time points . The naïve forecast model does not require any modeling exercises or computations, it simply returns the current value as future predictions, and thus . It's that simple. Even for this simple task, we will use the naïve function from the forecast package and ask it to provide the forecast for the next 25
observations with h=25
:
>co2_naive <- naive(co2_sub...