EVALUATION METRICS

• Closed-loop error: Test on the same data that was used to train the system.
  
  
• Open-loop error: Test on a held-out set.
  
  
• Cross-validation: Divide the data into various train and held-out sets and average the open loop error of each train and held-out set.
  
  
• Tree averaging: Build multiple trees using cross-validation and average the parameters of the trees