import numpy as np

import quapy as qp
from protocol import UPP
from quapy.method.aggregative import KDEyML

qp.environ["SAMPLE_SIZE"]=500


def datasets():
    for dataset_name in qp.datasets.UCI_MULTICLASS_DATASETS:
        yield qp.datasets.fetch_UCIMulticlassDataset(dataset_name)


for dataset in datasets():
    train, test = dataset.train_test
    test_gen = UPP(test, repeats=500)

    print(f"testing KDEy in {dataset.name}")
    for b in np.linspace(0.01, 20, 20):
        kdey = KDEyML(bandwidth=b, random_state=0)
        kdey.fit(train)

        report = qp.evaluation.evaluation_report(kdey, protocol=test_gen, error_metrics=['ae', 'rae', 'kld'], verbose=True)
        print(f'bandwidth={b}')
        print(f'MAE={report["ae"].mean()}')
        print(f'MRAE={report["ae"].mean()}')