I do not have a good theoretical background in machine learning, therefore will ask a maybe naive question, nonetheless, it's important for implementing my learning algorithm.
I'm having a question about dropping levels in machine learning classifiers. As I'm having about 20 nominal varaibles as my features, and each feature having 20 levels. Right now the learning speed of the gradient boosting machine with these features are pretty slow and the classification accuracy is also not as good as I want it to be.
My question is, does dropping levels in each of the nominal variable based on MCA may help with the learning speed and classification accuracy.
