STOPPING

• Stopping rules are used to prevent trees from overfitting the data

  
  

• General stopping rules

  
  

  • Node is pure -- all data belong to the same class

    
    

  • Impossible split -- all data have the same attribute values

    
    

• Other stopping rules

  
  

  • Minimum observed data -- a node with fewer observed examples than a predefined minimum value is turned into leaf

    
    

  • Maximum depth -- no further splitting of a node is allowed when the maximum tree depth is reached