Decision Tree (Regression)


  1. For each feature, sort the column and take the mean of two adjacent rows
  2. Use that mean as the node decision factor, i.e., age < 7, mean is 7
    1. For the decision, split them into two parts
    2. Average of each part's member is the predicted value
  3. Calculate sum of squared residual for each of the points for each node
  4. Take the one tree with smallest residual value
  5. If the number of candidate is less than threshold (usually 20) then exit
  6. Go to step 1