Error while hyperparameter tuning using BayesSearchCV for CatBoostClassifier (Multiclass classification)

Question

I am trying to tune hyperparameters for Multiclassclassification with CatBoostClassifier but getting below error.

ValueError: multiclass format is not supported

My target variable contains (0,1,2,3)

Please check below code that I have implemented.

  accuracy = make_scorer(accuracy_score, greater_is_better=True, needs_threshold=True)
skf = StratifiedKFold(n_splits=5, shuffle=True, random_state=42)

clf = CatBoostClassifier(thread_count=2, loss_function='MultiClass', eval_metric='Accuracy', leaf_estimation_method='Newton',  od_type = 'Iter', verbose= False)

# Defining your search space
search_spaces = {'iterations': Integer(10, 2000),
                 'depth': Integer(1, 12),
                 'learning_rate': Real(0.01, 1.0, 'log-uniform'),
                 'random_strength': Real(1e-9, 10, 'log-uniform'),
                 'bagging_temperature': Real(0.0, 1.0),
                 'border_count': Integer(1, 255),
                 'l2_leaf_reg': Integer(2, 30)}

# Setting up BayesSearchCV
opt = BayesSearchCV(clf,
                    search_spaces,
                    scoring=accuracy,
                    cv=skf,
                    n_iter=100,
                    n_jobs=1, 
                    return_train_score=False,
                    refit=True,
                    optimizer_kwargs={'base_estimator': 'GP'},
                    random_state=42)
opt.fit(X_train, y_train)

Please let me know if more detail is required regarding my question. Please help me to get this issue resolved. Thanks!

Answer 1

Try adding objective = 'multi:softmax' to the classifier and also setting classes_count