topic Re: Weighted accuracy calculation in QlikView

Weighted accuracy calculation

sangeess21 — Wed, 25 Nov 2020 16:16:04 GMT

Hi All,

I'm working on Weighted accuracy and I'm stuck at this point.

The rolled up formula for weighted accuracy is correct. The error was on each individual item when you sum it up by product classification. But the problem is the weights are not applied at the lowest level, they are assigned only at the second level in hierarchy. Rollup formula is 1-((error/demand)*weight) and the formula for individual weighted accuracy calculation is (1-(error/demand))*weight.

The hierarchy is (Low to high):

Item

Fcst Product Classification - Weights are assigned at this level

Item Group

Re: Weighted accuracy calculation

sunny_talwar — Tue, 13 Dec 2016 17:23:09 GMT

This is the part which is the issue

If(Dimensionality() = 0,

1 -

Sum({<Flag_Type=>}Aggr((Sum({<Flag_Type=>}Aggr(fabs(Sum(FORECAST_QTY)-Sum({<Flag_Type=>}Demand)),Item)))/Sum({<Flag_Type=>}Demand)*

Pick(Match(only({<Flag_Type=>}[Fcst Product Classification]), 'High Revenue', 'Commodity', 'High Growth Potential', 'EOS Pending', 'Build to Order', 'Legacy') + 1, 0,

$(vHighRevenue), $(vCommodity), $(vHighGrowthPotential), $(vEOSPending), $(vBuildToOrder), $(vLegacy)),

[Fcst Product Classification])),

if( not SubStringCount(Concat('"'&_dimension&'"', '|'), '"Fcst Product Classification"'),

(1 -

Sum({<Flag_Type=>}Aggr(((Sum({<Flag_Type=>}Aggr(fabs(Sum(FORECAST_QTY)-Sum({<Flag_Type=>}Demand)),Item)))/Sum({<Flag_Type=>}Demand)),$(=Concat(DISTINCT '[' & _DIM & ']', ', ')&',[Fcst Product Classification]'))))*

Pick(Match(only({<Flag_Type=>}[Fcst Product Classification]), 'High Revenue', 'Commodity', 'High Growth Potential', 'EOS Pending', 'Build to Order', 'Legacy') + 1, 0,

$(vHighRevenue), $(vCommodity), $(vHighGrowthPotential), $(vEOSPending), $(vBuildToOrder), $(vLegacy)),

(1 -Sum({<Flag_Type=>}Aggr(((Sum({<Flag_Type=>}Aggr(fabs(Sum(FORECAST_QTY)-Sum({<Flag_Type=>}Demand)),Item)))/Sum({<Flag_Type=>}Demand)),$(=Concat(DISTINCT '[' & _DIM & ']', ', ')))))*

Pick(Match(only({<Flag_Type=>}[Fcst Product Classification]), 'High Revenue', 'Commodity', 'High Growth Potential', 'EOS Pending', 'Build to Order', 'Legacy') + 1, 0,

$(vHighRevenue), $(vCommodity), $(vHighGrowthPotential), $(vEOSPending), $(vBuildToOrder), $(vLegacy))))

I guess what are your expected output when you have only Division as a dimension?

Re: Weighted accuracy calculation

sangeess21 — Tue, 13 Dec 2016 17:31:48 GMT

I want weighted accuracy to return results on all possible dimension selections. The above expression works only when Fcst Product Classification is selected in addition to other dimensions.

Re: Weighted accuracy calculation

sunny_talwar — Tue, 13 Dec 2016 17:51:05 GMT

Not sure, but can you check the attached. It would be helpful if I knew what are the numbers that you expect to see.... but I guess this will help

Re: Weighted accuracy calculation

sangeess21 — Wed, 14 Dec 2016 21:32:05 GMT

Thanks Sunny The expression works as expected.