4 ms·
Some techniques, e.g. random forest, give variable importance indicators for free. If you can test it out, give it a go - don't have to use the random forest as
by idunning 12y ago
Some techniques, e.g. random forest, give variable importance indicators for free. If you can test it out, give it a go - don't have to use the random forest as the final model.
- ma2rten 12y agoYou can use correlation or the coefficients of a linear model iff the features on the same scale. Another method is to train a model leaving out each feature once, then you see how much accuracy drops.