More Predictive Analytics: Microsoft Excel

4. Initializing Forecasts

In This Chapter

Setting Initial Values

Getting an Initial Forecast with Backcasting

Getting an Initial Forecast with Optimization

Initializing with Regression

Deciding to Use Polynomial Regression

If you want to avoid making a complete guess at forecasting the first few values of a time series, you have to have a baseline. That baseline should be a series of actual observations, derived from some sort of empirical process. The baseline gives you something to work with.

In practice, you arrive at that ideal after you’ve had some experience with whatever it is you want to forecast. Perhaps that is corporate revenues; perhaps it’s a hospital’s daily admissions for a particular virus; it might be something less weighty, ...