What are the advantages and disadvantages of a GBM model?

Advantages:

  • High accuracy
  • Tends to work well for imbalanced classification problems
  • Little time required for feature pre-processing
  • Does not require distributional assumption (only need to specify loss function)
  • Handles both numeric and categorical data types
  • Can extract feature importances and use base concept of decision tree for interpretation

Disadvantages:

  • Requires some computing power and time spent in parameter tuning
  • No direct interpretation such as regression equation is produced in output