|
Hello All, I'm trying to learn a binary classifier from few training examples (~30 positive, 30 negative). I have 10 predictors i.e. dimension of my feature vector is 10 (all are important predictors). How should I approach to counter the curse of dimensionality here? I heard that bootstrapping data in such cases is a good idea but can someone please shed a light on the process? Also, which classifiers people generally prefer in such scenario? I'm thinking of training decision trees (may be RF too). Thanks in advance! |
|
Decision tree & RF are multi-class classifier by its nature, but ur problem is a binary classification problem. So, you cand do it with a simple binary svm classifier. I'm not sure why decision trees can't be used for binary classification problem. Yes, I could use SVM or LR but I have one additional constraint if I use weight based classifiers. I want all of my weights to be +ve. Now, the standard implementation of the classifiers don't have that constraint and I will have to write my own code for that :(
(Jan 09 '14 at 20:59)
Ankit Jain
|
|
Decision tree & RF are multi-class classifier by its nature, but ur problem is a binary classification problem. So, you cand do it with a simple binary svm classifier. |
