Which algorithms are good for Multiclass classification problems?

classification
multi-class

#1

What algorithms are good for multi-class classifcation problem, when we have no. of features <10 ?
Please help


Which algorithms are good for Multi class classification problems?
#2

@Soham_Goswami,

Multi class classification problem can be solved in different ways:

  1. Create a binary variable for each class and predict them individually as binary classification after that combine the results but it is not the right choice if we have high number of classes because it takes good processing time. This binary classifier for multiclass can be used with one-vs-all or all-vs-all reduction method. Here you can go with logistic regression, decision tree algorithms.

  2. You can go with algorithms like Naive Bayes, Neural Networks and SVM to solve multi class problem.

  3. You can also go with multi layers modeling also, first group classes in different categories and then apply other modeling techniques over it.

Hope this help!

Regards,
Imran


#3

From my experience Ensemble algorithms tend to work well in many cases such as Latent Drichlet Allocation, Boosting, Bagging, Maximum Entropy and Nueral Net…


#4

@Soham_Goswami For a small testing of what works and what algorithm doesn’t you can use Weka, Its a tiny software where you can test before even writing a line of code

https://www.cs.waikato.ac.nz/ml/weka/