Hi Experts,
My client (Pharma) uses various Univariate forecast models (11,21,31,35,41,80,94)-Alpha .30, Beta.30, gamma.30, Sigma 1.25, Period 12, wherever in forecast model applicable.
No Trend Dampening/ No outliner correction as history correction done every month with same respective stat model too.
History Horizon is 36 months and forecast period is 18 months.
Now , when forecast model is changed to Auto Model (Alpha .30,Beta .30, Gamma .30 Sigma 1.25 and seasonal periods=12.
But due to insufficent history for few sku's. i.e SKU's with history less than for 12-24 months, Forecast is genrating for 2-3 months in a year instead of for 18 months.
Forecast accuracy is bases on weighted average of forecast of M1 (.4), M2(.4) AND M3(.2)
If We run stat forecast for sku X where history is available for 15 months, Stat is genrated for random months like Jan'17,Apr'17, Aug'17,Dec'17.
What kind of solution can be implemented to overcome this issue for SKU's with history less than 24 months.
I was looking for LIKE Model, as we have similar sku's with 36 months history.
Wis to grab some expert inputs before proceeding.
Regards
Add comment