|
I have five independent classifiers. I have a dataset, and run each classifiers using the same dataset,so I can get five different results. Now I want to combine the results from the five classifiers in order to improve the accuracy. Which methods are better and which software can solve this kind of problem? Thank you! |
|
Here are some things to try:
Now I think Mixture of expert and stacked generalized can deal with this problem. Do you know which software have these functions?
(Oct 25 '11 at 19:14)
yunfeiyu
At least with stacked generalization, you're probably going to need to implement it yourself. It's more of a method of how you feed information to the classifier than some blackbox algorithm.
(Oct 26 '11 at 01:24)
Kirk Roberts
1
I'd also suggest doing gdahl's options in order - I've had plenty of success with simple voting classifiers, and for actual implementations its rare I need to go past that.
(Oct 26 '11 at 04:58)
Robert Layton
Alas, yunfeiyu, I don't know of software for you, and as Kirk mentioned it may be hard to find such software.
(Nov 07 '11 at 02:01)
gdahl ♦
I am also looking for an stacking implementation, preferably in Python. Unfortunately, It seems that only Weka has a stacking implementation in the meta package.
(Mar 23 '12 at 22:00)
Xinfan Meng
|
|
Though I did not read it entirely, the book Pattern Classification Using Ensemble Methods seems to treat this subject in depth. |
