Options

What will be the suitable model for predicting a binominal class (Solved)

stephen5huostephen5huo Member Posts: 5 Contributor II
edited November 2018 in Help
Hi, a quick questions for the model selection in Rapid Miner, I’ve got a dataset with around 100 attributes and a class label with binomial value (1 or 0). Could anyone give some suggestion of what type of data mining model will be suitable for the prediction? Is there any example/tutorial in the forum with relate to this? Many thanks!

Answers

  • Options
    MariusHelfMariusHelf RapidMiner Certified Expert, Member Posts: 1,869 Unicorn
    Hi Stephen,

    for that problem virually any model which supports binominal labels can be used, e.g. Naive Bayes, SVMs etc. We have a lot of useful introductory videos linked on our website at http://rapid-i.com

    Best regards,
    Marius
  • Options
    stephen5huostephen5huo Member Posts: 5 Contributor II
    Sorry jus caught this, was in holiday :)
    SVM actually works well for my dataset, many thanks!
  • Options
    MariusHelfMariusHelf RapidMiner Certified Expert, Member Posts: 1,869 Unicorn
    Nice to hear that. In that case you may want to optimize the C parameter of the SVM to get even better results, and also try another kernel type, most notably the rbf/radial kernel. In that case, also the gamma-parameter must be optimized. For both parameters you should try values between 0,00001 and 10000. Have a look at the Optimize Parameters (Grid) operator and the logarithmic scale option on it.

    Best,
    Marius
  • Options
    stephen5huostephen5huo Member Posts: 5 Contributor II
    Hi Marius,

      Many thanks for the help! could you also share some tips to find the optimize value? as I am just basiclly manual try different C value to find the optimize one, any trick to speed up this process?
Sign In or Register to comment.