Hi
I encoded my dependent data with one-hot encoding, now I have this variable in several columns let say 11. Can I perform a Random Forest for feature selection for many targets?
To find the best-fitted data for my dependent.
Thanks, please do not comment just with a link.
I would like to suggest you some useful links https://www.ijcai.org/Proceedings/16/Papers/233.pdf
https://www.ingentaconnect.com/content/ben/cmc/2019/00000026/00000026/art00003
https://www.ingentaconnect.com/content/ben/cmc/2019/00000026/00000026/art00003
I'm not sure what you need feature selection for here, but to answer your question about multi-class random forest:
Yes this is possible, not sure what language you're using but this can rather easily be implemented in sk-learn in python. There is a blog post here:
https://www.codementor.io/@agarrahul01/multiclass-classification-using-random-forest-on-scikit-learn-library-hkk4lwawu
I don't think you need to one-hot encode the data yourself, I believe python will do that for you if you feed in a target vector of the various classes, no matter how many classes there are.
Would look at potentially other classification techniques as well. Multi-class classification is an interesting beast and you may find that another multi-class algorithm performs better. You could use decision trees, KNN, MLP, LDA, QDA, or just vanilla multiclass logistic regression. These approaches are all listed here (for sk-learn):
https://scikit-learn.org/stable/modules/multiclass.html
As a side note, I also hate when people just post links. Defeats the purpose of a scientific COMMUNITY.
Anyway, good luck, let me know if this helps.
- Badges
- Science topic
- Similar topics
- Analytical Chemistry
- Column