Input ExampleSet does not match the training ExampleSet

User46772

Hello, I am following the book "Rapid Miner for the Masses", and, trying to run the Neural Net model (Chapter 11, paragraph 5), I get an error, Attribute do not match - The input ExampleSet does not match the training ExampleSet. Missing Attribute 'Years_Pro'. Does someone have a hint? Thanks, Luis. 

hbajpai

Hey @User46772,

I am not familiar with the textbook but the error implies that Years_Pro attribute was in your training dataset but is absent from the apply model set. You can try to drop the attribute from the training set and it should not have this error.

BalazsBaranyRM

Hi,

you are applying the Windowing operator on the data for building the model, but you don't do the same for the data set you put into Apply Model.

You really need to do the same preprocessing for applying the model. You can probably simply copy the Windowing operator from the upper execution branch to the lower one.

Regards,
Balázs

hbajpai

Hey @User46772,

I am not familiar with the textbook but the error implies that Years_Pro attribute was in your training dataset but is absent from the apply model set. You can try to drop the attribute from the training set and it should not have this error.

User46772

Solved,but in a different way, I created the attribute in the dataset that did not have it, because this attribute had a role inthe Set Role operator. 

User13405

I'm the author of Data Mining for the Masses, so I can shed a little more light on this.

This was a minor error in the data set for the 1st Edition of Data Mining for the Masses (2012), which I suspect is the edition of the book you are using. There are two ways to fix the problem. The first is to open the Chapter 11 Training data CSV file in a text editor or spreadsheet application and just change the variable name ‘Years_Pro’ to be ‘Years_Exp’. Save the change and close the file, then re-import or re-connect to the data in RapidMiner and re-run the process. The error will go away once the variable names between the Training and Scoring data sets match.

The other way to fix the problem is to download the Chapter 11 data sets from the websites for either the Second or Third editions of the book. Here are the URLs for each of those editions:

https://sites.google.com/site/dataminingforthemasses/se

https://sites.google.com/site/dataminingforthemasses3e/

The Chapter 11 data sets have been simplified from the version I used in the first edition. There are fewer attributes in the latter data sets. Please let me know if you have any additional trouble on this or other topics in Data Mining for the Masses. Note that you can get newer editions of the book on Amazon or at myeducator.com if you want to.

Matt North
mnorth@uvu.edu

akaplan

Type your comment

BalazsBaranyRM

Hi,

you are applying the Windowing operator on the data for building the model, but you don't do the same for the data set you put into Apply Model.

You really need to do the same preprocessing for applying the model. You can probably simply copy the Windowing operator from the upper execution branch to the lower one.

Regards,
Balázs

anms

Hi,

I am newbie & having the same problem too. I don't understand what does that mean & what should I do. Hope someone could guide me.

Thank you.

BalazsBaranyRM

Hi @anms,

you're doing the Nominal to Numerical before applying the SVM on the left side so the support vector machine model is built on changed data. However, the same processing is not being done on the right side: You put the original structure into Apply Model. Of course there will be missing attributes. 

The recommended way for this is using Group Models. Insert a Group Models operator into the Training process. Connect the "pre" output of Nominal to Numerical to the first input of Group Models. Connect the "mod" output of SVM to the second input. Connect the Group Models output to the "mod" output of the Training subprocess.

This will create a combined model that applies Nominal to Numerical and SVM to the original dataset both when training and testing. 

Regards,
Balázs

anms

Many thanks BalazsBarany for your explaination. I manage to get it when using split validation operator.

However, when I'm try to apply feature selection operator (fwd selection/backward elimination), this message appear. 

I have activated the debug mode, but still the process failed. 


The process flow seems okay since all operators have green ticks; but Rapidminer could not produce the result output. 


Hope you could help me to solve this matter. Thank you very much.

MartinLiebig

Hi,

please check your rapidminer-studio.log. It should have more information.

Best,

Martin