Missing Attribute Values in Train. Ex.

• Blood-Test-Result
  1. standard methodology from Statistics is throw data away
  2. assign missing value to the most common value at node
  3. alternatively, assign missing value to the most common value at node for examples with the same target value
  4. assign probability to each possible value, estimated by frequencies at node
• latter tack, can be subdivided again later in the tree
• same approach can be used to classify examples