Custom Evaluation Metrics XGBoost: Precision and f1_score

custom_eval_metrics_xgboost.py

def xgb_precision(proba_y: np.ndarray, dataset: xgb.DMatrix) -> Tuple[str, float]:
  '''returns binary classification precision using 0.5 threshold.
     proba_y: 1x2 shape or binary classification probabilities
     dataset: xgboost DMatrix
  '''
  y = dataset.get_label()
  tresh_func = np.vectorize(lambda x: 1 if x> 0.5 else 0)
  pred_y = tresh_func(proba_y)
  return 'clf_precision', precision_score(y, pred_y)

def xgb_f1(proba_y: np.ndarray, dataset: xgb.DMatrix) -> Tuple[str, float]:
  '''returns classification f1_score using 0.5 threshold.
     proba_y: 1x2 shape or binary classification probabilities
     dataset: xgboost DMatrix
  '''
  y = dataset.get_label()
  tresh_func = np.vectorize(lambda x: 1 if x> 0.5 else 0)
  pred_y = tresh_func(proba_y)
  return 'clf_f1', f1_score(y, pred_y)