The time plot in Figure 19 . 15 describes actual quarterly sales for a department store over a six-year period (data are available in DepartmentStoreSales.csv, data courtesy of Chris Albright).

a. Which of the following methods would not be suitable for forecasting this series?

• Moving average of raw series

• Moving average of deseasonalized series

• Simple exponential smoothing of the raw series

• Double exponential smoothing of the raw series

• Holt-Winters exponential smoothing of the raw series

b. The forecaster was tasked to generate forecasts for four quarters ahead. He therefore partitioned the data such that the last four quarters were designated as the holdout period. The forecaster approached the forecasting task by using multiplicative Holt-Winters exponential smoothing. The smoothing parameters used were α = 0.5, β = 0.0001, γ = 0.2 with multiplicative seasonality.
i. Run this method on the data.
ii. The forecasts for the holdout set are given in Table 19 2. Compute the MAPE values for the forecasts of Q21 and Q22.

c. The fit and residuals from the exponential smoothing are shown in Figure 19 . 16. Using all the information thus far, is this model suitable for forecasting Q21 and Q22?

Transcribed Image Text:

Sales 100,000 90,000- 80,000 70,000- 60,000- 50,000- 0 10 Quarter 15 201 25