CrossValidation is often used in conjunction with so-called (hyper)parameter tuning. What are hyperparameters? These are the various knobs that you can tweak on your machine learning algorithm. For example, these are some parameters of the Random Forest classifier:
- Number of trees
- Feature subset strategy
- Impurity
- Maximal number of bins
- Maximal tree depth
Setting these parameters can have a significant influence on the performance of the trained classifier. Often, there is no way of choosing them based on a clear recipe--of course, experience helps--but hyperparameter tuning is considered as black magic. Can't we just choose many different parameters and test the prediction performance? Of course, we can. This feature ...
