🎉 🎉 RAPIDMINER 9.10 IS OUT!!! 🎉🎉

Download the latest version helping analytics teams accelerate time-to-value for streaming and IIOT use cases.


brute force feature selection

gutompfgutompf Member Posts: 21  Maven
edited June 2019 in Help
I was working with Backward elimination and Forward e. feature reduction processes and it works fine. Now I want to run BruteForce process, but it take all my memory (8 Gb, I am on W7 64-bit) and crash. It sayz me that I should find other computer with more memory or to reduce data. I would like to know"
1. if there are not some other possibilities, I mean some settings of windows or RapidMiner?
2. I have LIBSVM in BruteForce as learner - cannot help choosing some other learner?
3. Or when I will reduce data, what should be reduced - columns or rows?


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

    how many attributes do you have? Since you create the powerset of all attributes, it might need quite a lot of memory. If you want to reduce the data, you can try both dimensions - but probably reducing the number of columns will have the greater impact on this problem.
    Concerning the learner: in many domains the learner in not that important for feature selection, but if a feature set works good for one learner, it also works well for other learners. So maybe you can try Naive Bayes as feature learner and have a look if it generates a similar set of features compared to the SVM (you may want to check this on a smaller example set first).

Sign In or Register to comment.