Example

.. Examples documentation master file.

Examples

Multi-class classification

import the cgb

  >>> from cgb import cgb_clf, cgb_reg
  >>> import sklearn.datasets as dts
  >>> from sklearn.model_selection import train_test_split

You may fit the model for a dataset with n>2 class labels. Although it works perfectly for binary problems as well.

  >>> X, y = dts.make_classification(n_samples=100, n_classes=3,
                               n_clusters_per_class=2,
                               random_state=1, n_informative=4)

Split the data. Here to show the instruction of the model we consider a simple split. We suggest choosing the proper splitting method with regard to your dataset.

  >>> x_train, x_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=1)

import C_GradientBoostingClassifier from the cgb. For classification, you should use deviance as the loss function. You may leave other hyperparameters with default values.

  >>>  model = cgb_clf(max_depth=5,
                subsample=1,
                max_features='sqrt',
                learning_rate=0.1,
                random_state=1,
                criterion="squared_error",
                loss="log_loss",
                n_estimators=100)

Fit the model with 100 trees

  >>> model.fit(x_train, y_train)

Performance of the model. Returns the accuracy

  >>> model.score(x_test, y_test)

Multi-output regression

You may fit the model for a dataset with n>2 outputs. Although it works perfectly for simple regression too.

  >>> X, y = dts.make_regression(n_samples=100, n_features=100, n_targets=3)

import cgb_reg from the cgb. For regression, you may leave the loss.

  >>> model = cgb_reg(learning_rate=0.1,
                subsample=1,
                max_features="sqrt",
                loss='ls',
                n_estimators=100,
                max_depth=3,
                random_state=1)

Fit the model with 100 trees.

  >>> model.fit(x_train, y_train)
  

Performance of the model. Returns the RMSE for n_outputs.

  >>> model.score(x_test, y_test)

predict

Print predicted values of the trained model.

  >>> model.predict(x_test)

Returns np.array(N-features, n_classes/outputs)

staged_predict

Print predicted values for each base leaner

  >>> pred = np.zeros((y_test.shape[0], model.n_estimators))
  >>> for i, pred_i in enumerate(model.staged_predict(x_test)):
        pred[:, i] = pred_i

score_stage

Use the staged_predict to return the score for stages

  >>> from sklearn.metrics import accuracy_score
  >>> test_score = np.empty((model.n_estimators))
  >>> for i in range(model.n_estimators):
        test_score[i] = accuracy_score(y, pred_cgb[:, i])