Part $($a$)$ Use the $4-$period moving average method to forecast for these data and for the first $4$

weeks after week $16.$ Compute MSD$/$E$,$ MAD, MAPE, and Bias. Plot the forecasting

errors e$($t$)$ in Excel to see any patterns in the errors.

Part $($b$)$ Use the first$-$order exponential smoothing model with $\backslash$alpha $=0\mathrm{.}3$ to forecast for these data

and for the first $4$ weeks after week $16.$ Compute MSD$/$E$,$ MAD, MAPE, and Bias.

Plot the forecasting errors e$($t$)$ in Excel to see any patterns in the errors.

$($Assume F$(0)=$ A$(1))$

Part $($c$)$ Use the second$-$order exponential smoothing model with $\backslash$alpha $=\backslash$beta $=0\mathrm{.}3$ to forecast for

these data and for the first $4$ weeks after week $16.$ Compute MSD$/$E$,$ MAD, MAPE,

and Bias. Plot the forecasting errors e$($t$)$ in Excel to see any patterns in the errors.

$($Assume F$(0)=$ A$(1),$ T$(0)=0\mathrm{.}0))$

Part $($d$)$ Use Winters$$s multiplicative exponential smoothing model with $\backslash$alpha $=\backslash$beta $=\backslash$gamma $=0\mathrm{.}3$ and N$=4046.,+100$

to forecast for these data and for the first $4$ weeks after week $16.$ Compute MSD$/$E$,$

MAD, MAPE, and Bias. Plot the forecasting errors e$($t$)$ in Excel to see any patterns in

the errors.

$($Assume F$($N$)=$ Average of the first N A$($i$)$ values, T$($N$)=0\mathrm{.}0))$

Part $($e$)$ Same as Part $($d$)$ except $\backslash$alpha $=0\mathrm{.}7,\backslash$beta $=\backslash$gamma $=0\mathrm{.}1$ and N $=4.$

Which technique is best and why? Suggest modification$($s$)$ to the above models that

may make more sense for this situation.