Gradient Boost (Classification)


  1. Get initial guess, but unlike Gradient Boost (Regression), this time we will take log(odds)
  2. Get the residual by assuming one class as 1.0 and other class 0.0
  3. Fit a tree
  4. $current_guees = initial_guess * current_prediction$
  5. Convert current guess to log_odds
  6. go to step 2
  7. Stop when predetermined number of estimator is reached