def fit_model(train_pool, test_pool, **kwargs):
    model = CatBoostClassifier(
        max_ctr_complexity=5,
        task_type='CPU',
        iterations=10000,
        eval_metric='AUC',
        od_type='Iter',
        od_wait=500,
        **kwargs
    )return model.fit(
        train_pool,
        eval_set=test_pool,
        verbose=1000,
        plot=False,
        use_best_model=True)

def classification_rate(y, y_pred):
    return np.sum(y==y_pred)/len(y)

def get_oof(n_folds, x_train, y, x_test, cat_features, seeds):    ntrain = x_train.shape[0]
    ntest = x_test.shape[0]
    oof_train = np.zeros((len(seeds), ntrain, 3))
    oof_test = np.zeros((ntest, 3))
    oof_test_skf = np.empty((len(seeds), n_folds, ntest, 3))    test_pool = Pool(data=x_test, cat_features=cat_features)
    models = {}    for iseed, seed in enumerate(seeds):
        kf = StratifiedKFold(
            n_splits=n_folds,
            shuffle=True,
            random_state=seed)
        for i, (train_index, test_index) in enumerate(kf.split(x_train, y)):
            print(f'\nSeed {seed}, Fold {i}')
            x_tr = x_train.iloc[train_index, :]
            y_tr = y[train_index]
            x_te = x_train.iloc[test_index, :]
            y_te = y[test_index]
            train_pool = Pool(data=x_tr, label=y_tr, cat_features=cat_features)
            valid_pool = Pool(data=x_te, label=y_te, cat_features=cat_features)model = fit_model(
                train_pool, valid_pool,
                loss_function='MultiClass',
                random_seed=seed
            )
            oof_train[iseed, test_index, :] = model.predict_proba(x_te)
            oof_test_skf[iseed, i, :, :] = model.predict_proba(x_test)
            models[(seed, i)] = modeloof_test[:, :] = oof_test_skf.mean(axis=1).mean(axis=0)
    oof_train = oof_train.mean(axis=0)
    return oof_train, oof_test, models

balanced accuracy: 0.6703822994494413
classification rate: 0.8198316498316498

balanced accuracy: 0.6549535670689709
classification rate: 0.8108249158249158