2 A naïve prediction of the future
This chapter covers
- Defining a baseline model
- Setting a baseline using the mean
- Setting a baseline using the mean previous window of time
- Setting a baseline using the previous time step
- Implementing the naïve seasonal forecast
In chapter 1, we covered what time series are and how forecasting a time series is different from a traditional regression task. Also, we learned the necessary steps to build a successful forecasting project, from defining a goal to building model, deploying it, and updating it as new data is collected. Now, we are ready to start forecasting a time series.
We will first learn how to make a naïve prediction of the future, which will serve as a baseline. The baseline model is a trivial solution that uses heuristics, or simple statistics, to compute a forecast. Developing a baseline model is not always an exact science. It will often require some intuition that we gain by visualizing the data and detecting patterns that can be used to make predictions. In any modeling project, it is important to have a baseline, as we will use it to compare the performance of the more complex models we will build down the road. The only way to know that a model is good, or performant, is to compare it to a baseline.