Error: Model Comparison: Tool #3: Error in names(scores) <- paste(score.field, "_", y.leve

Hi, I am Angela. I have been working on developing a logistic regression model and using the model comparison tool to compare models. I have been troubled by errors in the model comparison tool. I have tried to change the variables and punctuations but it doesn't help to solve the errors.

The error: Model Comparison: Tool #3: Error in names(scores) <- paste(score.field, "_", y.levels, sep = "") :

Any help would be appreciated!

Please find attached the workflow and input file. Thanks.

Copy of ASM logistic.yxmd

Screenshot 2021-06-07 at 3.36.53 AM.png

Predictive Analysis

Accepted answers

apathetichell

Something like this? It has both in an excel sheet. Could you mark one of these posts as a Solution so that the next time someone is struggling with their R Models hopefully they can get some pointers in how to troubleshoot their workflow/data.

asm_logistic_score.xlsx

All comments

CristonS

hi @AngelaLilan you forgot to include your input file, which the Community will need to help troubleshoot. The easiest way is to package your workflow as a yxzp (Options> Export workflow). Thanks!

AngelaLilan

Sorry that I forgot to include the input file.

I really appreciate any help as I am clueless now. Millions of Thanks!

Dataset A_Final_2020-2021-1-1.xls

AngelaLilan

Hi, thank you for your help.

But after adding the 'union' tool, both of the 'model comparison' and 'score' are still having the same errors.

I have tried to change the predictor variables but it does not help.

Also, if I would like to change the target variable from 'responded to email campaign' to downloaded discount voucher for purchase', there will be the same errors for 'stepwise' and 'score'.

Could you help me to check again?

Furthermore, may I ask how can I use R directly to make samples?

Sorry as I am new to Alteryx, the questions may sound silly.

Thanks.

Copy of Copy of ASM logistic.yxmd

Screenshot 2021-06-08 at 3.14.50 PM.png

apathetichell

Yeah I got your workflow to run perfectly with the original variable but changing the target variable isn't working for me. I'll come back to it later today and see if I can get it to work. I've seen scenarios where you need to change the tools out to get them to work... for sample split with continuity for the dependent variable you can just this in an an R tool:

change "DependentVariable" to whatever column you are using. This assumes that you are feeding in data with static "setseed" information and "sampleratio" information (on my system this is in a not ready for primetime macro which adds/adjust these) so just hardcode them if you don't have as variables. As you can see r tool out 1 will be your train/evaluation data and r tool out 2 will be your test/evaluation data.

library(caTools)

data<-read.Alteryx("#1", mode="data.frame")
set.seed(data$setseed[1])

data$sampleratio[1]

spl=sample.split(data$DepedentVariable,SplitRatio=data$sampleratio[1])
data_train <-subset(data,spl==TRUE)
data_test <-subset(data,spl==FALSE)

data_train = subset(data_tran, select = -c(setseed,sampleratio) )
data_test = subset(data_test, select = -c(setseed,sampleratio) )

write.Alteryx(data_train,1)
write.Alteryx(data_test,2)

AngelaLilan

Thank you.

I have tried to add an R tool but I guess something is missing so that the data is not run successfully. I have tried to add the below code as yours in R tool but is it that I just copy and paste? Or I have to change something?

library(caTools)

data<-read.Alteryx("#1", mode="data.frame")
set.seed(data$setseed[1])

data$sampleratio[1]

spl=sample.split(data$DepedentVariable,SplitRatio=data$sampleratio[1])
data_train <-subset(data,spl==TRUE)
data_test <-subset(data,spl==FALSE)

data_train = subset(data_tran, select = -c(setseed,sampleratio) )
data_test = subset(data_test, select = -c(setseed,sampleratio) )

write.Alteryx(data_train,1)
write.Alteryx(data_test,2)

Thanks.

Copy of Copy of ASM logistic.yxmd

Screenshot 2021-06-08 at 4.17.44 PM.png

apathetichell

sorry let me clean it up:

library(caTools)

data<-read.Alteryx("#1", mode="data.frame")
set.seed(YOUR SET SEED HERE)

spl=sample.split(data$(NAME OF YOUR DEPENDENT VARIABLE HERE), SplitRatio=(SAMPLE SPLIT RATIO HERE)
data_train <-subset(data,spl==TRUE)
data_test <-subset(data,spl==FALSE)

write.Alteryx(data_train,1)
write.Alteryx(data_test,2)

Try that. I'm carrying over values in columns in a set format which obviously your data won't have.

apathetichell

2021-06-08 (1).png

And here's the version of the new dependent variable - I left some notes on the canvas explaining some of things I did and why.

revised asm logistic - new dependent.yxmd

AngelaLilan

Thank you very much for your help.

But what do you mean by adding a new version of the same tool when using a model to fix an error on the canvas?

Following that, I have added a decision tree, another model comparison and lift chart to the workflow.
But the same errors show up again

Error: Decision Tree (35): Decision Tree: The R.exe exit code (1) indicated an error.

A new error showed up too. It leads to the lift chart not showing any result.

Error: Decision Tree (35): Decision Tree: Error in relevel.factor(actualValues, ref = ref) :

May I know how can I fix it and why?

Thanks.

Screenshot 2021-06-09 at 12.39.02 AM.png

Copy of revised asm logistic - new dependent.yxmd

apathetichell

adding a decision tree to it gave me an error because it wanted a complexity parameter. I added one and got the following:

2021-06-08 (2).png

2021-06-08 (3).png

2021-06-08 (4).png

2021-06-08 (5).png

apathetichell

The tree tool really doesn't like "West" in the Region category. I had the hyperparameters minimum value set to 20 and with:

Region Count
Central 272
East 127
Midlands 358
North 277
North-East 84
North-West 189
South 60
South-East 138
South-West 92
West 36

in your total data set - the training version didn't have over 20 instances so it didn't want to create any splits with West as a qualifier. At least - that's how I'm reading it.

AngelaLilan

So should I change the number for 'The minimum number of records needed to allow for a split'? Or what should I change?

I have tried to change 'The minimum number of records needed to allow for a split'? Or what should I change? ranges from 5-15 but none of it seems to solve the error.

Thanks.

AngelaLilan

Ok thanks.

But I have tried to run the workflow again today but it said :

Error: R (36): AlteryxProcessManager::Start: Timed out in ConnectNamedPipe after 5000 milliseconds

for all the models. May I know what is going on?

Thanks.

Screenshot 2021-06-10 at 4.49.05 AM.png

apathetichell

Hi!

I have no idea - as I haven't seen that error - maybe someone from technical support can help? Here's what some forum people reported on that issue:

https://community.alteryx.com/t5/Alteryx-Designer-Discussions/Error-in-ConnectNamedPipe/td-p/81262/page/2

there's also the disconnect the R tool and then reconnect it option. So most Alteryx tools just kind of do their things and exist in a relative vacuum - but that's not true for R, Python, R-based tools, and a few others. So they get more hiccups.

maybe @CristonS can help?

AngelaLilan

Hi Please. Thanks.

Perhaps can you send me the file again and see if I can run it or not? Millions of thanks.

Copy of revised asm logistic - new dependent.yxmd

apathetichell

This was the version I did with the (working) tree tool.

revised asm logistic - new dependent.yxmd

AngelaLilan

It does not work 😞

Would you mind capturing me the report for logistic regression, stepwise and model comparison? As well as making a decision tree and model comparison.

Sorry I know I am asking a bit more but this workflow is really important to me as it is an urgent work that I have to analyse within hours. Millions of thanks,

apathetichell

Do you want me to run the score with one the models too?

modelcompare.xlsx

model compare.pdf

CristonS

Great work, @apathetichell! ConnectNamedPipe errors are quite often environmental, so I recommend reaching out to our Support team. They'll know the right questions to ask in terms of whitelisting, anti virus software (McAfee can cause problems), etc. Cheers!

AngelaLilan

It will be much appreciated if you can also run the score for the models.

Thank you!

apathetichell

Do you prefer the Tree model? I think the stepwise is currently hooked up. also - I could run both if you want- and post the output as .xlsx. Also you wanted the model run with all the data- no? I'm no expert in these things but with the thresholds for your Tree set so low (to include the West Region values), I'd want to scrutinize the model for overfitting.

For future R workflows, it might be worth reinstalling R and even turning off anti-virus when running some of the workflows. As @CristonS mentioned - R is very machine/environment specific. You can also do test runs of your data in Rstudio to see if its flagging the same errors.

apathetichell

asm_logistic_score.xlsx

AngelaLilan

I prefer running both Decision Tree and Stepwise models. I don’t think all the data should be run as I saw that the errors are high. And as I remembered, the root node errors are high, not sure how can I reduce the errors.

Yes you could scrutinize the model for overfitting.

Thanks.

apathetichell

the .xlsx has both with the stepwise and the decision tree run on all the data and labeled as such. If you want it just run on the eval/test data let me know and I can set that up too. Also I could render if if you want more readable excel data.

AngelaLilan

Thank you.

could you run the evaluation/test data and render it? Thanks.

but I think for overall, the models generate exceptionally large errors. I guess it’s because of the data being insufficient?

apathetichell

could be - could be that there's a variable that isn't being captured. or the time frame isn't long enough for a pattern to form. updated version - first version wasn't set up to show decimals....

originallm_testdata.xlsx

rajanalad

Hi Angela,

Looks like the error is due to the "(" round brackets that you've placed around the variable name and the split ratio. Try running the workflow again after removing the brackets and it should work.

Hope this helps.

Regards,

Dhivya