https://communities.sas.com/t5/SAS-Viya/Is-it-possible-to-replicate-the-use-of-decisions-node-in-Model/m-p/952164#M2673 <P>Hi</P><P>&nbsp;</P><P>In some of our models created in SAS Enterprise Miner we are using the decisions node to adjust the probabilities coming out of the model. Is it possible to replicate this feature in Model Studio in some way?</P> Fri, 29 Nov 2024 10:23:01 GMT RichardHorgen 2024-11-29T10:23:01Z https://communities.sas.com/t5/SAS-Viya/Is-it-possible-to-replicate-the-use-of-decisions-node-in-Model/m-p/952164#M2673 <P>Hi</P><P>&nbsp;</P><P>In some of our models created in SAS Enterprise Miner we are using the decisions node to adjust the probabilities coming out of the model. Is it possible to replicate this feature in Model Studio in some way?</P> Fri, 29 Nov 2024 10:23:01 GMT https://communities.sas.com/t5/SAS-Viya/Is-it-possible-to-replicate-the-use-of-decisions-node-in-Model/m-p/952164#M2673 RichardHorgen 2024-11-29T10:23:01Z https://communities.sas.com/t5/SAS-Viya/Is-it-possible-to-replicate-the-use-of-decisions-node-in-Model/m-p/952186#M2674 <P><A href="https://communities.sas.com/t5/SAS-Communities-Library/Model-Studio-for-SAS-Enterprise-Miner-Users-Part-1/ta-p/915913" target="_blank" rel="noopener">Model Studio for Enterprise Miner Users: Part 1</A></P> <P><A href="https://communities.sas.com/t5/SAS-Communities-Library/Model-Studio-for-SAS-Enterprise-Miner-Users-Part-2-Data/ta-p/922279" target="_blank" rel="noopener">Model Studio for Enterprise Miner Users: Part 2, Data</A></P> <P><A href="https://communities.sas.com/t5/SAS-Communities-Library/Model-Studio-for-SAS-Enterprise-Miner-Users-Part-3-Let-s-get/ta-p/944134" target="_blank" rel="noopener">Model Studio for Enterprise Miner Users: Part 3, Let’s get philosophical</A></P> <P><A href="https://communities.sas.com/t5/SAS-Communities-Library/Model-Studio-for-SAS-Enterprise-Miner-Users-Part-4-Partitioning/ta-p/951352" target="_blank" rel="noopener">Model Studio for SAS Enterprise Miner Users: Part 4, Partitioning Data</A></P> <P>&nbsp;</P> <P>I'm not sure which feature of the Enterprise Miner Decisions node you are looking for in Model Studio ,</P> <P>but I guess it's about adjusting the posterior probabilities for the real priors (maybe you also used EM's target profiler?) to make them more "honest". You could call this "calibration".</P> <P>&nbsp;</P> <P>The answer to your question is maybe in this Communities Library Article :</P> <P>Implementing Out-of-Time Testing in Model Studio<BR />Last update: ‎01-AUG-2023<BR />Updated by: SAS Employee Sharad_Saxena</P> <P><A href="https://communities.sas.com/t5/SAS-Communities-Library/Implementing-Out-of-Time-Testing-in-Model-Studio/ta-p/887246" target="_blank" rel="noopener">https://communities.sas.com/t5/SAS-Communities-Library/Implementing-Out-of-Time-Testing-in-Model-Studio/ta-p/887246</A></P> <P><STRONG>...&nbsp;The score code contains a section titled&nbsp;Adjust Posterior Probabilities. ...</STRONG></P> <P>&nbsp;</P> <P><SPAN>I don't think (not sure though!) you can specify the real priors -- by typing them in -- somewhere in your Model Studio project.</SPAN></P> <P><SPAN>What does this mean? --&gt; Well ... if you don't use the event-based sampling facility of Model Studio for having a better balance between outcome categories because the data you give to Model Studio are already under-sampled or over-sampled before (in other words ... the very initial data seen by Model Studio do not contain an accurate representation of the true proportions between the outcome categories) ... then you are going to have to adjust the posterior probabilities yourselves to account for the real priors. This can be done with some simple formulas.</SPAN></P> <P>&nbsp;</P> <P>Ciao, Koen</P> Fri, 29 Nov 2024 21:58:15 GMT https://communities.sas.com/t5/SAS-Viya/Is-it-possible-to-replicate-the-use-of-decisions-node-in-Model/m-p/952186#M2674 sbxkoenk 2024-11-29T21:58:15Z