xgboost - XGBRegressor loss functio custom

Question

I want to custom loss function to quantile loss(pinball loss) in XGBRegressor

I use this code

def xgb_quantile_eval(preds, dmatrix, quantile=0.2):
    labels = dmatrix.get_label()
    return ('q{}_loss'.format(quantile),
            np.nanmean((preds >= labels) * (1 - quantile) * (preds - labels) +
                       (preds < labels) * quantile * (labels - preds)))


def xgb_quantile_obj(preds, dmatrix, quantile=0.2):
    try:
        assert 0 <= quantile <= 1
    except AssertionError:
        raise ValueError("Quantile value must be float between 0 and 1.")

    labels = dmatrix.get_label()
    errors = preds - labels

    left_mask = errors < 0
    right_mask = errors > 0

    grad = -quantile * left_mask + (1 - quantile) * right_mask
    hess = np.ones_like(preds)

    return grad, hess

And I build model like this

def XGB(q, X_train, Y_train, X_valid, Y_valid, X_test):    
    # (a) Modeling  
    model = XGBRegressor(objective=xgb_quantile_obj, alpha=q,
                     n_estimators=10000, bagging_fraction=0.7, learning_rate=0.027, subsample=0.7)                 
                     
    model.fit(X_train, Y_train, eval_metric = xgb_quantile_eval, 
          eval_set=[(X_valid, Y_valid)], early_stopping_rounds=300, verbose=500)

    # (b) Predictions
    pred = pd.Series(model.predict(X_test).round(2))
    return model, pred

But I got an error

models_2, results_2 = XGB(0.5, X_train_1, Y_train_1, X_valid_1, Y_valid_1, X_test)
results_2

• AttributeError: 'numpy.ndarray' object has no attribute 'get_label'

I am not sure if I am doing well. Please help me

Answer 1

Oh I catch the error, I have to change the sequence between preds and dmatrix in xgb_quantile_obj and change dmatrix to labels