Baselines for comparisons: NAIVE and Window Regression

[mdancho84]

Baseline Methods

An important way to compare cutting-edge performance is to use a baseline model, these are simple models that organizations are accustomed to using (e.g. simple moving average or naive models). We can use these to showcase how much better high-performance methods like stacking, ensembling, and better algorithms (e.g. XGBoost) can do.

Comparison models

There are currently two types of comparison models:

1. Window Regression (window_reg) - This can be used to showcase how a moving average, weighted average, or even simple seasonal models would perform.
2. NAIVE Regression (naive_reg) - This can be used to perform NAIVE (Pick most recent observation) and Seasonal NAIVE (replicate most recent seasonal sequence).