how can i do statistical analysis of 3 models result in RapidMiner?

pradee_sivarampradee_sivaram Member Posts: 4 Contributor I
edited December 2018 in Help

Hello all,

I have 3 different models comparative results (precision, recall, accuracy, micro and macro precision-recall) in Java implementation result. now I want to do a statistical analysis of those 3 models. please help me how to do it in RapidMiner.

tanks in advance

Answers

  • Telcontar120Telcontar120 Moderator, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,635 Unicorn

    What exactly do you mean by a "statistical analysis of the 3 models"? Do you mean a performance comparison?  If so, "Performance to Data" may help.  If you are looking for more classical statistics, you might be better off with an R script. 

    Brian T.
    Lindon Ventures 
    Data Science Consulting from Certified RapidMiner Experts
  • pradee_sivarampradee_sivaram Member Posts: 4 Contributor I

    thanks for your reply.

    I have evaluated experiment result value of system identified(using Java), human identified result(manual), true positive, true negative, false positive and false negative values for my 3 models(ie 3 methods).
    using this result I want to identify which model gives the best result in statistical analysis.

    (i don't know about statistical analysis, rapid miner, R, and i know only java) 

    so pls give the solution for this

  • Thomas_OttThomas_Ott RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,761 Unicorn

    There are summary statistics if you use the Statistics View after your run your results.

     

    For Machine learning models you want to use the Performance operators to measure how well you classify your results. For example, you could use LogLoss, Area Under Curve (AUC), or even Kappa. I would use those. 

  • pradee_sivarampradee_sivaram Member Posts: 4 Contributor I

    thanks for your reply sir.

    please tell me, sir, whether we can use this kappa statistics for comparing 3 models result and/or classify one model observed and expected results?

    because I need to find which model gives the best result.

    suppose if I used kappa metric and found its value like model1=69, model2 =74, model3=78

    then can I conclude models 3 is the best model?

     

    please reply me

    thanks in advance

     

Sign In or Register to comment.