"Correlation Matrix"

simsim Member Posts: 18 Learner I
edited May 2019 in Help
I am trying to conduct a correlation matrix on some data. However the results do not include a correlation matrix, but rather a table with two columns where all of the attributes only in one column. I have used the "nominal to binomial", "correlation matrix" and "select weights" operators. 
Do you know what I am doing wrong?
Tagged:

Best Answer

Answers

  • Telcontar120Telcontar120 Moderator, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,308   Unicorn
    If you can post your XML it would be easier to troubleshoot :-)
    But from your description, it sounds like you might be using Weight by Correlation, which only looks at the correlation between attributes and the defined label.  If you want the full correlation matrix you need to use the Correlation Matrix operator instead.

    Brian T.
    Lindon Ventures 
    Data Science Consulting from Certified RapidMiner Experts
  • simsim Member Posts: 18 Learner I
    Hi Telcontar120, 
    Thank you for such a quick response! I have removed the "select weights" attribute, but am still facing the same error. I would upload the XML file, but don't know how to (I'm new to rapidminer), sorry!

    Do you know if there's anything else that I can try?

  • simsim Member Posts: 18 Learner I
    Thank you mschmitz!!! That definitely helped!! I now have my results in the form of a correlation table.
    All of the categories within my attributes are now listed as individual attributes, is there anyway for this to be adjusted? 
    Thank you once again! 
  • mschmitzmschmitz Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, University Professor Posts: 2,253  RM Data Scientist
    Hi,
    pearson correlation is not defined for nominal types. Thus they can't be in.

    BR,
    Martin
    - Head of Data Science Services at RapidMiner -
    Dortmund, Germany
  • simsim Member Posts: 18 Learner I
    is there an operator than can be used to convert the data so it can be included?
  • mschmitzmschmitz Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, University Professor Posts: 2,253  RM Data Scientist
    Well, i would take a measure which can handle this. i.e Weight by Gini Index.
    - Head of Data Science Services at RapidMiner -
    Dortmund, Germany
  • simsim Member Posts: 18 Learner I
    does the weight by ginni index convert the data?
  • simsim Member Posts: 18 Learner I
    Hi Weight by Ginni Index did not work for me- is there anything else that I can use?
  • simsim Member Posts: 18 Learner I
    edited January 2019
    Hi mschmitz, hope you're well!
    Just wondering if there was an update?
  • simsim Member Posts: 18 Learner I
    Hi Martin, 

    Just wondering if you've seen my above message?

  • mschmitzmschmitz Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, University Professor Posts: 2,253  RM Data Scientist
    Hi @sim ,
    i would go for something like the attached one. but please keep in mind that this is only defined for not necesarrly normalized w.r.t correlation.
    <?xml version="1.0" encoding="UTF-8"?><process version="9.1.000">


    - Head of Data Science Services at RapidMiner -
    Dortmund, Germany
    David_A
Sign In or Register to comment.