Error in min function in summarize

[Amit]

Hi,

My input data looks like this:

Here, we can see that the output should choose record number 1,4,6,7 as the output

But, the output that I am getting is:

For each SKU Input the correct MAPE value is taken but the corresponding model name is ARIMA everywhere instead it should be ETS for the first and last

Am I doing something wrong in the way I am using the summarize tool?

[jdunkerley79 · Accepted solution]

You need to do this in two steps.

First find the Min MAPE for each SKU and then use a join tool to join back to the input to select the Model with the same MAPE and SKU.

In its current form the summarise pick the minimum sting (hence ARIMA) for each SKU.

[jack_morgan · Accepted solution]

Morning Amit!

So expand on the post above, Min/Max on text (sting, vsting, wstring, etc.) looks for the min or max value based on alpha. Arima has a lower letter than ETS, thus Arima was selected every time. If you want to get your MIN MAPE and still have the detial of each line, you would need to run a separate data stream to get your MIN and then Join on the SKU and MIN_MAPE to the original data source (your table in this case). Results that come out of the join are those that you are looking for. I've attached a very basic workflow to help demonstrate what you need to do to get your desired result.

Jack

[jack_morgan]

I thought of another method which would be to use the sort and unique tool. Sort by the criteria that would place your desired record at the top and then use the uniqe tool for the sku only.

See attached file (it has both potential workflows and the same result; 1,4,6,7).

Best of luck!

[Amit]

Thank you so much for your response guys, finally I understood why it was showing ARIMA all the time :)