About WendyCzika

WendyCzika · ‎10-01-2018

I'd suggest contacting SAS Technical Support for this issue: https://support.sas.com/en/technical-support.html

WendyCzika · ‎09-27-2018

You need to set your target to have the level of Binary in your Input Data node.

WendyCzika · ‎09-26-2018

You should be able to use the Merge node in EM for this. Just be sure to set the Merging property to Match, select Account as a By Ordering variable, set Overwrite Variables property to No, and any variables that are named the same in both data sets need to be flagged with No for Overwrite Variable in the Edit Variables window.

WendyCzika · ‎09-20-2018

This post gives a list of other variables with reserved names or prefixes in EM: https://communities.sas.com/t5/SAS-Communities-Library/Scoring-Series-Part-2-SAS-Enterprise-Miner-Scoring-Output/ta-p/223842 You can also take a look at this post about global metadata to see how to programmatically change variable roles, levels, etc. instead of manually: https://communities.sas.com/t5/SAS-Communities-Library/Tip-Defining-Global-Metadata-for-SAS-Enterprise-Miner-Projects/ta-p/223835

WendyCzika · ‎09-20-2018

Do the variables marked as Residual start with the prefix "R_" ?

WendyCzika · ‎09-13-2018

Those properties in Data Partition just control what reports are included in the results of that node, so shouldn't affect the Decision Tree node.

WendyCzika · ‎09-12-2018

I'm not sure I understand your statement: "I believe this is because my target variable is a ratio that ideally I want to be close to 1, but the tree is trying to find how to maximize the target." The Decision Tree is trying to come up with rules based on your inputs for predicting your target, that is, the splitting rules that create the most significant difference in the target values of the resulting branches.

WendyCzika · ‎09-11-2018

By default, TANH is the activation function for each hidden layer, but you can change it to SIN, IDENTITY, or TANH for each layer: hidden 1000 / act=sin; hidden 500 / act = tanh;

WendyCzika · ‎09-10-2018

I think I know what you're asking about, but not 100% sure so if you could include a screenshot, that would be helpful. If it's about how sometimes you only see part of the rule in the HP Tree results in the Tree plot, with a "..." at the end, in that case you might have to look at the score code or Node Rules to see the entire rule.

WendyCzika · ‎08-14-2018

If you are using the Advanced advisor in your Input Data node, then variables are rejected if they have excessive missing values (>50% by default) and excessive class values (>20 by default). You can turn off the setting of variables to rejected or you can change the cutoffs used in the Advanced Options property of the Input Data node when you select Advanced for the Advisor property or you can customize the Advanced options when going through the Data Source Wizard (step 4).

WendyCzika · ‎08-10-2018

Here are 2 posts that might help you: https://communities.sas.com/t5/SAS-Communities-Library/Tip-How-Can-I-Apply-HP-Forest-Score-Code-to-Distributed-Data/ta-p/221493 https://communities.sas.com/t5/SAS-Communities-Library/Scoring-with-Analytic-Stores/ta-p/362845

WendyCzika · ‎08-10-2018

I think it's because the values for Suco_Fr are all missing in your score data set, so it can't calculate a true prediction based on the regression model, so just uses the mean of the target from the training data for the prediction.

WendyCzika · ‎08-10-2018

If you look at the results of your Regression, do you see any effects being included in the model? I'm guessing it is resulting in an intercept-only model so that is why you are getting all the same predictions.

WendyCzika · ‎08-03-2018

If using the Neural Network node in Enterprise Miner, you can follow that with a Score node with the properties "C Score" and/or "Java Score" set to Yes, then you can use either C or Java to score.

WendyCzika · ‎07-26-2018

The only other suggestion I have is to make sure the Hide Rejected Variables property of the Metadata node is set to No and an option other than None is selected for the Combine Rule property. If it still doesn't work, please contact SAS Technical Support: https://support.sas.com/en/technical-support/contact-sas.html

Online Status	Offline
Date Last Visited	a week ago