Forecasting: ARIMA vs. Exponential Smoothing (R and Tableau)
Forecasting is a tricky subject. Within limits, we can only forecast depending on certain conditions being true. That's why I start with "What can be forecast?" followed by an ARIMA model in R vs. an Exponential Smoothing Forecast in Tableau (default parameters). What can be forecast? The predictability of an event or a quantity depends on several factors including: 1. How well we understand the factors that contribute to it. 2. How much data is available. 3. Whether the forecast can affect the thing we are trying to forecast.
For example, forecasts of electricity demand can be highly accurate because all three conditions are usually satisfied. On the other hand, when forecasting currency exchange rates, only one of the conditions is satisfied: there is plenty of available data. However, we have a limited understanding of the factors that affect exchange rates. Models can be simple (e.g. exponential moving average) to complex (e.g. ANOVA, DL). Short-term forecasts: Scheduling personnel, production, and transportation. Medium-term forecasts: Determine future requirements (materials, personnel) Long-term forecasts: Strategic planning of several years.
The ARIMA sample is taken from the book "Forecasting: Principles and Practice" by Prof. Rob J. Hyndman, probably the foremost expert on forecasting (in R).
In Tableau (see below), I've used the very same dataset and forecast period with Tableau's default forecast which is based on exponential smoothing.
My personal conclusion is that a "simple" exponential smoothing for forecasting is a powerful alternative to the ARIMA model. Best of all, as Data Scientists, our primary job is to support users. With Tableau's built-in forecasting model, we probably can cover 80% or more of use cases without falling back to more complicated models such as ARIMA.