topic Re: Forecast error MAPE in SAS Forecasting and Econometrics

Forecast error MAPE

Ghabek — Sat, 01 Feb 2020 04:56:52 GMT

I’m running visual forecasting in viya with retail data
10k records and region and state hierarchy

I’m getting WMAPE and MAPE values in the hundreds ....500, 600, etc

Anyone know how to get them to be normal less than 100?

This is urgent for customer meeting

I’m a SAS partner

Thx

919-610-3700

Re: Forecast error MAPE

imvash — Sat, 01 Feb 2020 05:58:12 GMT

Can you run the demand classification pipeline and see how many series are considered intermittent?

Re: Forecast error MAPE

Ghabek — Sat, 01 Feb 2020 14:02:12 GMT

Anyway you can call me to figure this out with me pls?

Thanks

9196103700

Re: Forecast error MAPE

Ghabek — Sat, 01 Feb 2020 20:41:46 GMT

hers the data

if you could take look

im here to try and figure this out!

Hierarchy is Region and State

Dependent is total sales

time is order date by week

Re: Forecast error MAPE

imvash — Mon, 03 Feb 2020 16:29:24 GMT

The problem with your data is the high number of missing values in your data when you aggregate it at the weekly level. For example, the followings are the most problematic series (the number in parentheses are missing values out of series with 209 observations):

West : Wyoming (208)
Central : North Dakota (207)
East : West Virginia (207)
East : Maine (206)
East : District of Co (205)
Central : South Dakota (204)
East : Vermont (203)
West : Montana (201)

Note: I used the following code to aggregate data at weekly level:

proc import datafile = '~\ctg_retail_demo.xlsx'
	out = retail_data replace;
run;

proc sort data = retail_data out = retail_data_sorted;
	by state region Order_Date;
run;

proc timedata data = retail_data_sorted out = out;
	id Order_Date interval = week;
	var Total_Sales / ACCUMULATE=TOTAL;
	by state region;
run;