Prediction

LeMarc

Hello,

I was wondering how to predict the existence of missing values? I do not mean to predict the values itself but merely is there a missing value or not within the field.

Thank you

varunm1

Hello @LeMarc

Do you mean you want to make a classification task with Missing (Yes or No) as a label column?

You can use generate attributes operator to create a new attribute that has Missing status. This will provide you with True or False in that new attribute. If the example has a missing value in a particular field, it will display the status as TRUE else FALSE.

[Deleted User]

Hello @LeMarc,

Take a look at this link please. With declare missing value operator you can declares the specified values of the selected attributes as missing values.

 https://docs.rapidminer.com/latest/studio/operators/cleansing/missing/declare_missing_value.html

Also may be you need this one

https://docs.rapidminer.com/latest/studio/operators/cleansing/missing/replace_missing_values.html

There is a question which is similar to your question and the link is this:

https://community.rapidminer.com/discussion/comment/61107#Comment_61107


I hope this helps
mbs

LeMarc

Thank you @mbs

those operators are familiar to me as well.

I was meaning how to predict if a field within a label attribute is missing or not. So that would be a binominal classification task. Missing - Yes or No.

Is that possible?

LeMarc

@ varunm1 Thank you.

I tried to generate a new attribute that has missing status. That should be based on all existing columns. But it is not possible to select several attributes for the missing function at once? Quote " Error: The function missing must have 1 argument but  has x".

And also I would like to cluster the examples with missing values withing a single cluster. But it seems most of the available algorithms cant handle missing values anyway. Is there a solution?

LeMarc

Ok found the solution to my first answer on how to select several attributes for the missing function at once. But an answer to the second question would be helpful!

varunm1

I would like to cluster the examples with missing values withing a single cluster. But it seems most of the available algorithms cant handle missing values anyway. Is there a solution?

@LeMarc Yep, Not all algorithms handle missing values. General procedures are to either impute missing values or remove examples with missing values. Few algorithms that can handle clustering with missing values are Support Vector Clustering and Random Clustering.

You can also right-click on any operator and click show operator info to see what types of data they support.

LeMarc

@varunm1 Thank you for the answer