community
cancel
Showing results for 
Search instead for 
Did you mean: 

Alteryx designer Discussions

Find answers, ask questions, and share expertise about Alteryx Designer.
Upgrade Alteryx Designer in 10 Steps

Debating whether or not to upgrade to the latest version of Alteryx Designer?

LEARN MORE

Random Forest model output - MSE available?

Highlighted

Hello all, 

 

I am trying to extract data from the model output of various predictive tools, but mainly Random Forest. 

After learning a bit of R, I can extract the IncNodePurity using the 'importance' call like so:

 

model.data <- read.Alteryx("#1")
the_obj <- unserializeObject(as.character(model.data$Object[1]))

the_summary <- summary(the_obj)
the_importance <- importance(the_obj)

write.Alteryx(as.data.frame(the_summary), 1, TRUE)
write.Alteryx(as.data.frame(the_importance), 2, TRUE)

Importance gives the Node Purity but MSE is more accurate and I would prefer to output this instead. Only, I can't find a reliable way to call %IncMSE from the R tool. Does anyone know how to extract this, and more generally is there a place with comprehensive documentation of these? The 'Tool Mastery' pages are not detailed enough!

 

As a side note, It seems bizarre that the tools would output PNG of the data in the Reports but have a not have a formatted, tabular output as well. It would certainly save me some time. Is this in the pipeline?

Alteryx
Alteryx

Hello @samstafford,

 

Since the Forest Model tool uses the RandomForest package, it might be a good idea to check out the documentation for that package: randomForest

Random Forest Package.png

 

As for pulling the %IncMSE, here are a few articles that may help you further develop your custom code:

https://stackoverflow.com/questions/27918320/what-does-negative-incmse-in-randomforest-package-mean - importance(rf,type=1)

https://stats.stackexchange.com/questions/162465/in-a-random-forest-is-larger-incmse-better-or-worse

https://www.quora.com/How-do-we-calculate-variable-importance-for-a-regression-tree-in-random-forest...

 

Also, if you think it would be a valuable feature for the predictive tools to output Reports as a data stream, here is a post on our ideas page that you can star/comment on:

https://community.alteryx.com/t5/Alteryx-Designer-Ideas/Output-Predictive-Report-Tables-as-Data-Stre...

 

Hoss Carroll
Customer Support Engineer
Alteryx

Hello Hoss,

 

Thanks for your suggestions.

After looking through the documentation for randomForest and the stackexchange posts I am still unable to extract the values. Your sources explicitly say that IncMSE and IncNodePurity are both returned when calling rf$importance or importance(rf), but Alteryx using the R tool only returns IncNodePurity. It's almost as if Alteryx is using a non-standard randomForest package?

 

For the record the output says I'm using randomForest 4.6-14

Alteryx Certified Partner
Alteryx Certified Partner

You could try using the model comparison tool, which is available on the gallery.

 

https://gallery.alteryx.com/#!app/Model-Comparison/56bbd3013df7da08b8fcd00a

 

The purpose of the tool is to compare the output of multiple models, but it can be used for just a single model too, it will return the RMSE, MAE, MPE and MAPE values.

 

https://help.alteryx.com/2018.2/Model_Comparison.htm

 

Ben

Labels