SVM vs Random forest




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?




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.