0
$\begingroup$

I need an library, or something that is already done for SVM and Random Forest algorithms. Can you give me some ideas? I don't have experience and I don't know what to choose.

The restriction of my classification problem is: 27 dimensions, 9 classes, 50.000 entries in the training set, 150.000 in test set.

$\endgroup$
2
  • 1
    $\begingroup$ Try scikit-learn. $\endgroup$ Apr 11, 2015 at 8:22
  • 1
    $\begingroup$ Package e1071,randomForest,caret in R are pretty easy.The difficult part comes when adjusting model parameters whichever library you use $\endgroup$
    – D.Castro
    Apr 11, 2015 at 9:01

1 Answer 1

2
$\begingroup$

I, too, would suggest the 'caret' package in R

You can built a lot of models and compare their performances

http://topepo.github.io/caret/training.html

By the way, usually the ratio of the training set to the test set is a bit higher than that you have.

Let have a look at this discussion: https://stackoverflow.com/questions/13610074/is-there-a-rule-of-thumb-for-how-to-divide-a-dataset-into-training-and-validatio

$\endgroup$
4
  • 1
    $\begingroup$ But few people have the luxury of such a large data set. $n_{train} = 50000 : p = 27$ is a pretty comfortable training set size, and 1.5e5 test cases is nice as well (assuming these are independent cases...) $\endgroup$ Apr 11, 2015 at 16:42
  • $\begingroup$ I understand your point, but in the machine learning community, usually the training set is suggested to be larger than the test set, not the other way around. $\endgroup$ Apr 12, 2015 at 18:14
  • $\begingroup$ There are several discussions on this point: stats.stackexchange.com/questions/23331/… $\endgroup$ Apr 12, 2015 at 18:15
  • $\begingroup$ Also see this: quora.com/… $\endgroup$ Apr 12, 2015 at 18:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.