SVM vs Random forest

svm
classification
random_forest

#1

Hi,

Which algorithm out of Support Vector Machine and RandomForest should we use when we have to build a model for a classification problem? In which situations does SVMs perform better and in which does RandomForests?

Thanks.


#2

@Mark,

In general, you couldn’t say that which one is better than the other, both methods are promising. It depends on data and its distribution. Below I am listing some of comparison scenario that can help you to choose the significant technique (but this can not be rule to decide)

  • Random Forest is faster compare to SVM respect to processing time although the performance
    of each method is based upon the data and optimal input parameters.

  • If your output variable has two class and your data is reasonably clean and outlier free, structural risk minimization is a powerfull approach and I would go with SVM. In a many class case and data gas outliers then I would go for the Random Forest.

  • Random forest is a black box, you are not aware about the internal splits/ rules so it is difficult to deploy in manual intervention project.

Regards,
Imran