Algorithm 2: Supervised Feature Classification

Input: Training data named X, training label named Y, test data named X_test Output: The predict result of X_test 1. kfold_data = StratifiedKFold (folds_num = 5) 2. misjudgedX = [] 3. misjudgedY = [] 4. for train_index, val_index in kfold_data.split(X, Y) 5.  tra_x, val_x, tra_y, eval_y = X[train_index], X[val_index], Y[train_index], Y[val_index] 6.  if len(misjudgedX) ! = 0 7.   tra_x = vstack ([tra_x, misjudgedX]) 8.   tra_y = hstack([tra_y, misjudgedY]) 9.  end if 10.  model = lightgbm.train(tra_x, tra_y) 11.  val_pred = model.predict(val_x) 12.  misjudgedX = val_x[argwhere(val_y ! = val_pred)] 13.  misjudgedY = val_y[argwhere(val_y ! = val_pred)] 14.  test_pred.append(model.predict(X_test)) 15. end for 16. final_pred = [] 17. for line in test_pred 18.  final_pred.append(argmax(bincount(line))) 19. end for 20. return final_pred