Chapter 1: Combining Forecasting Methods: Predicting Quarterly Sales in 2019 for Motorola Solutions

Often there is more than one forecast method of a given time series data source. The dilemma exists when considering which of these possible forecasting models should be chosen to produce the most accurate result. The simplest method for combining several forecasts is to average them but some assert that it may not be advisable to use simple averages when there is a large difference among the variances of the errors.

Research by Makridakis et al. (1982) however, has shown that simple averages produce more accurate forecasts than weighted averages when used with many time series. The concept of using weighting based on the inverse sum of squared errors was developed by Bates and Granger (1969).