17.3 Advantages and Disadvantages of SVM

Advantages

• Performs well at classifying non-linear data
• Optimizing margins can help reduce the overfitting of data and allow for capacity control
• Learning without a local minima
• There are many kernels (transformations) that could be used to fit the data unlike any other algorithm
• Often provides sparse solutions
• Performs well on data sets that have many attributes, even if there are relatively very few cases on which to train the model

Limitations

• Choice of the right Kernel

  • The number of possible kernels is infinite and can make it hard to choose the right one
  • Most software uses a few kernels that generalize to many situations, but no kernel generalizes to every situation

• Can be computationally intensive

  • Algorithms can be complex
• Can overfit the model