def for_each_fold(fold, folds, data, labels, model, error_function):
(x_train, y_train), (x_test, y_test) = partition_data(data, labels, fold, folds)
model.fit(x_train, y_train)
y_pred = model.predict(x_test)
# Based on the error_function passed
if error_function is None: # if None calculate precision
error = precision_score(y_test, y_pred)
elif error_function == 'precision':
error = precision_score(y_test, y_pred)
if error_function == 'accuracy':
error = accuracy_score(y_test, y_pred)
elif error_function == 'recall':
error = recall_score(y_test, y_pred)
elif error_function == 'f1':
error = f1_score(y_test, y_pred)
else:
raise ValueError('%s error function is not defined.' % error_function)
return {'expected labels': y_test,
'predicted labels': y_pred,
'errors': [error]}
def test_accuracy_score_simple():
y_true_np = np.array([1, 0, 1, 0])
y_pred_np = np.array([1, 1, 0, 0])
y_true = torch.Tensor(y_true_np)
y_pred = torch.Tensor(y_pred_np)
acc = accuracy_score(y_pred, y_true)
assert acc == 0.5
def test(classifier,x_test,y_test):
prediction = classifier.predict(x_test)
print("Confusion Matrix for Decision tree Classofier Model given : ")
print(confusion_matrix(y_test,prediction))
print("Classification Report for given Decision tree Classisifer : ")
print(classification_report(y_test,prediction))
print("Accuracy Score : ",accuracy_score(y_test,prediction))
return
def run_1nn(points):
m = KNN(1)
m.train(points)
predicted = m.predict(points)
real = []
for point in points:
real.append(point.label)
print(accuracy_score(real, predicted))
def score(self, X, y):
"""
Returns the mean accuracy on the given test data (X, y).
"""
# Check that the trees have been fit
self._check_fit()
y_pred = pd.DataFrame(self.predict(X), columns=['y_hat'], index=y.index)
return metrics.accuracy_score(self.class_hierarchy, y, y_pred)
def get_best_model(X, y, clf, kf, clf_name, fitur, filename, show=False):
"""
fungsi untuk mendapatkan model terbaik dari hasil k-fold
return best_model: model terbaik, dengan tolak ukur gmean
parameter:
X = data per jenis fitur
y = label dari data
clf = object classifier
kf = object K-Fold
show = boolean untuk mencetak proses pencarian model terbaik
"""
performance_total = 0
best_fold_performance = -100
best_fold_index = -1
best_model = None
if show: print('\t\t', end='')
performances = []
performances.append(clf_name.upper() + '-' + fitur.upper())
for index, (train_index, test_index) in enumerate(kf):
X_train_fold, X_test_fold = X[train_index], X[test_index]
y_train_fold, y_test_fold = y[train_index], y[test_index]
clf_now = clf.fit(X_train_fold, y_train_fold)
pred = clf.predict(X_test_fold)
acc = round(accuracy_score(y_test_fold, pred) * 100, 2)
gmean_score = round(gmean(y_test_fold, pred) * 100, 2)
selected_metric_score = gmean_score
if selected_metric_score > best_fold_performance:
best_fold_performance = selected_metric_score
best_fold_index = index
best_model = clf_now
if show: print(selected_metric_score, end=' ')
performance_total += selected_metric_score
performances.append(selected_metric_score)
performance_avg = round(performance_total / 10, 2)
performances.append(performance_avg)
if show:
print(
'\n\t\tbest index: {}, best performance: {}, performance avg: {}\n'
.format(best_fold_index + 1, best_fold_performance,
performance_avg))
with open(filename, 'a', newline='') as file:
writer = csv.writer(file)
writer.writerow(performances)
return best_model
def evaluate(model, data_iterator, params, mark='Eval', verbose=True):
"""Evaluate the model on `steps` batches."""
# set model to evaluation mode
model.eval()
# id2tag dict
idx2tag = {idx: tag for idx, tag in enumerate(params.tags)}
true_tags = []
pred_tags = []
# a running average object for loss
loss_avg = utils.RunningAverage()
for input_ids, input_mask, labels in data_iterator:
# to device
input_ids = input_ids.to(params.device)
input_mask = input_mask.to(params.device)
labels = labels.to(params.device)
batch_size, max_len = labels.size()
# get loss
loss = model(input_ids, attention_mask=input_mask.bool(), labels=labels)
loss /= batch_size
# update the average loss
loss_avg.update(loss.item())
# inference
with torch.no_grad():
batch_output = model(input_ids, attention_mask=input_mask.bool())
# 恢复标签真实长度
real_batch_tags = []
for i in range(batch_size):
real_len = int(input_mask[i].sum())
real_batch_tags.append(labels[i][:real_len].to('cpu').numpy())
# List[int]
pred_tags.extend([idx2tag.get(idx) for indices in batch_output for idx in indices])
true_tags.extend([idx2tag.get(idx) for indices in real_batch_tags for idx in indices])
# sanity check
assert len(pred_tags) == len(true_tags), 'len(pred_tags) is not equal to len(true_tags)!'
# logging loss, f1 and report
metrics = {}
f1 = f1_score(true_tags, pred_tags)
accuracy = accuracy_score(true_tags, pred_tags)
metrics['loss'] = loss_avg()
metrics['f1'] = f1
metrics['accuracy'] = accuracy
metrics_str = "; ".join("{}: {:05.2f}".format(k, v) for k, v in metrics.items())
logging.info("- {} metrics: ".format(mark) + metrics_str)
# f1 classification report
if verbose:
report = classification_report(true_tags, pred_tags)
logging.info(report)
return metrics
def run_1nn(points):
m = KNN(1)
m.train(points)
predicts = []
real = []
for point in points:
predicts.append(m.predict(point)[0])
real.append(point.label)
print(accuracy_score(real, predicts))
def ques_one(points):
k = KNN(1)
k.train(points)
real = [0] * len(points)
predicted = [0] * len(points)
for i in range(len(points)):
real[i] = points[i].label
predicted[i] = k.predict(points[i])[0]
print("question 1 answer: ", accuracy_score(real, predicted))
def evaluate(model, data_iterator, params, mark='Eval', verbose=False):
"""Evaluate the model on `steps` batches."""
# set model to evaluation mode
model.eval()
idx2tag = params.idx2tag
true_tags = []
pred_tags = []
# a running average object for loss
loss_avg = utils.RunningAverage()
for _ in range(params.eval_steps):
# fetch the next evaluation batch
batch_data, batch_token_starts, batch_tags = next(data_iterator)
batch_masks = batch_data.gt(0)
loss = model((batch_data, batch_token_starts),
token_type_ids=None,
attention_mask=batch_masks,
labels=batch_tags)[0]
loss_avg.update(loss.item())
batch_output = model((batch_data, batch_token_starts),
token_type_ids=None,
attention_mask=batch_masks)[
0] # shape: (batch_size, max_len, num_labels)
batch_output = batch_output.detach().cpu().numpy()
batch_tags = batch_tags.to('cpu').numpy()
pred_tags.extend([[idx2tag.get(idx) for idx in indices]
for indices in np.argmax(batch_output, axis=2)])
true_tags.extend(
[[idx2tag.get(idx) if idx != -1 else 'O' for idx in indices]
for indices in batch_tags])
assert len(pred_tags) == len(true_tags)
# logging loss, f1 and report
metrics = {}
f1 = f1_score(true_tags, pred_tags)
metrics['loss'] = loss_avg()
metrics['f1'] = f1
metrics_str = "; ".join("{}: {:05.2f}".format(k, v)
for k, v in metrics.items())
logging.info("- {} metrics: ".format(mark) + metrics_str)
if verbose:
report_acc = accuracy_score(true_tags, pred_tags)
report = classification_report(true_tags, pred_tags)
logging.info(report_acc)
logging.info(report)
return metrics
def run_knn_kpoints(points, k):
m = KNN(k)
sum_ = 0
for _ in range(len(points)):
point = points[0]
points.remove(point)
m.train(points)
cl = m.predict(point)
sum_ += accuracy_score(point.label, cl)
points.append(point)
return sum_/len(points)
def run_knn1(points):
"""
a function for question 1
:param points: list of Point
"""
m = KNN(1)
m.train(points)
predicted = m.predict(points)
true_labels = []
for point in points:
true_labels.append(point.label)
print("accuracy_score for k=1:", accuracy_score(true_labels, predicted))
def cross_validation_whole():
X_train = []
X_test = []
y_train = []
y_test = []
indexes = [[0, 36], [37, 73], [74, 110], [111, 146],
[147, 182], [183, 216], [217, 253], [254, 290], [291, 328],
[329, 365]] #indexes of specified groups
res_accuracy = []
res_precision = []
for i in range(0, 10):
X_test = data[np.arange(indexes[i][0], indexes[i][1] + 1), :48]
X_test = X_test[:, [0, 1, 38, 39, 40, 41, 42, 43]] #F1+F12
X_test = X_test.astype(np.float64)
X_train = np.delete(data,
np.arange(indexes[i][0], indexes[i][1] + 1),
axis=0)[:, :48]
X_train = X_train[:, [0, 1, 38, 39, 40, 41, 42, 43]] #F1+F12
X_train = X_train.astype(np.float64)
y_test = data[np.arange(indexes[i][0], indexes[i][1] + 1), 48]
y_train = np.delete(data,
np.arange(indexes[i][0], indexes[i][1] + 1),
axis=0)[:, 48]
#clf = svm.SVC(kernel='rbf', probability=0, C=1).fit(X_train, y_train)
#clf = LogisticRegression(solver='liblinear', C=10).fit(X_train, y_train)
clf = RandomForestClassifier(criterion='gini',
n_estimators=10,
min_samples_leaf=1).fit(X_train, y_train)
#clf = DecisionTreeClassifier(criterion='entropy', max_features='auto', min_samples_leaf=2).fit(X_train, y_train)
y_pred = clf.predict(X_test)
res_accuracy.append(
metrics.accuracy_score(y_test, y_pred, normalize=True))
res_precision.append(
metrics.precision_score(y_test, y_pred, pos_label="1"))
res_accuracy = np.array(res_accuracy)
res_precision = np.array(res_precision)
print(res_accuracy.min())
print(statistics.median(res_accuracy))
print(res_accuracy.max())
print(res_precision.min())
print(statistics.median(res_precision))
print(res_precision.max())
print("max acc: " + str(res_accuracy.max()) + " min acc: " +
str(res_accuracy.min()) + " mediana acc: " +
str(statistics.median(res_accuracy)))
print("max prec: " + str(res_precision.max()) + " min prec: " +
str(res_precision.min()) + " mediana prec: " +
str(statistics.median(res_precision)))
def evaluate(self, darray, thr):
batch_index = 0
X_batch, P_batch, y_batch = self.get_batch(darray, self.batch_size,
batch_index)
y_pred = None
y_label = None
while len(X_batch) > 0:
num_batch = len(y_batch)
feed_dict = {
self.vocab_index:
X_batch,
self.props:
P_batch,
self.label:
y_batch,
self.first_level_lstm_dropout_p:
[1.0] * len(self.first_level_lstm_dropout),
self.deep_dropout_p: [1.0] * len(self.deep_dropout),
self.conv_pool_dropout_p: [1.0] * len(self.conv_pool_dropout),
self.second_level_lstm_dropout_p:
[1.0] * len(self.second_level_lstm_dropout),
self.train_phase:
False
}
batch_out = self.sess.run(self.out, feed_dict=feed_dict)
if batch_index == 0:
y_pred = np.reshape(batch_out, (num_batch, ))
y_label = np.reshape(y_batch, (num_batch, ))
else:
y_pred = np.concatenate(
(y_pred, np.reshape(batch_out, (num_batch, ))))
y_label = np.concatenate(
(y_label, np.reshape(y_batch, (num_batch, ))))
batch_index += 1
X_batch, P_batch, y_batch = self.get_batch(darray, self.batch_size,
batch_index)
pred = [1 if y_pred[i] > thr else 0 for i in range(len(y_pred))]
accuracy = metrics.accuracy_score(y_label, pred)
precision = metrics.precision_score(y_label, pred)
recall = metrics.recall_score(y_label, pred)
f1 = metrics.f1_score(y_label, pred)
return accuracy, precision, recall, f1
def on_batch_close(self, loss: torch.Tensor, np_probs: torch.Tensor,
targets: torch.Tensor):
# np_probs N*2*H*W targets: N*H*W
# targets = torch.zeros(size=np_probs.shape).scatter_(dim=1, index=targets.unsqueeze(dim=1).long(), value=1)
np_preds = torch.argmax(np_probs, dim=1).squeeze()
assert np_preds.shape == targets.shape
self.batch_num += 1
if not torch.isnan(loss):
self.metrics['loss'] += float(loss)
dice: torch.Tensor = metrics.dice_score(np_preds, targets)
if not torch.isnan(dice):
self.metrics['dice'] += float(dice)
# iou: torch.Tensor = metrics.iou_score(np_preds, targets)
# if not torch.isnan(iou):
# self.metrics['iou'] += float(iou)
acc: torch.Tensor = metrics.accuracy_score(np_preds, targets)
if not torch.isnan(acc):
self.metrics['acc'] += float(acc)
def step(self):
""" Epochs step, training and validation.
Return:
training_loss, validation_loss, accuracy, precision, recall
"""
# Training loop
batch_loss, batch_val_loss, batch_accuracy, batch_precision, batch_recall = [], [], [], [], []
for x_batch, y_batch in self.train_loader:
x_batch = x_batch.to(self.device)
y_batch = y_batch.to(self.device)
loss = self.train_step(x_batch, y_batch)
batch_loss.append(loss)
with torch.no_grad():
# Validation loop
for i, (x_val, y_val) in enumerate(self.val_loader):
x_val = x_val.to(self.device)
y_val = y_val.to(self.device)
self.model.eval()
yhat = self.model(x_val)
val_loss = self.criterion(yhat, y_val)
batch_val_loss.append(val_loss)
batch_accuracy.append(accuracy_score(np.argmax(yhat.cpu().detach().numpy(), axis=1), np.argmax(y_val.cpu().detach().numpy(), axis=1)))
batch_precision.append(precision_score(np.argmax(yhat.cpu().detach().numpy(), axis=1), np.argmax(y_val.cpu().detach().numpy(), axis=1)))
batch_recall.append(recall_score(np.argmax(yhat.cpu().detach().numpy(), axis=1), np.argmax(y_val.cpu().detach().numpy(), axis=1)))
# step lr scheduler using val_loss
if self.scheduler is not None:
self.scheduler.step(val_loss)
return [ torch.mean(torch.Tensor(batch_loss)),
torch.mean(torch.Tensor(batch_val_loss)),
torch.mean(torch.Tensor(batch_accuracy)),
torch.mean(torch.Tensor(batch_precision)),
torch.mean(torch.Tensor(batch_recall))
]
def generate_classification_perf(truths, pred_probs, multiclass=False):
"""Given truths, and predicted probabilities, generate ModelPerf object"""
pred_classes = np.round(pred_probs).astype(int)
with warnings.catch_warnings():
warnings.simplefilter("ignore")
retval = ClassificationModelPerf(
auroc=metrics.roc_auc_score(truths, pred_probs),
auroc_curve=metrics.roc_curve(truths, pred_probs)
if not multiclass else None,
auprc=metrics.average_precision_score(truths, pred_probs),
accuracy=metrics.accuracy_score(truths, pred_classes)
if not multiclass else None,
recall=metrics.recall_score(truths, pred_classes)
if not multiclass else None,
precision=metrics.precision_score(truths, pred_classes)
if not multiclass else None,
f1=metrics.f1_score(truths, pred_classes)
if not multiclass else None,
ce_loss=metrics.log_loss(truths, pred_probs, normalize=False) /
np.prod(truths.shape),
)
return retval
from bayes import NaiveBayes
from util import FileOperate
from util import train_test_split
from metrics import accuracy_score
# 运行这部分代码的时候，要将 playML 这个文件夹设置为源代码的根文件夹
if __name__ == '__main__':
# 1、加载数据，spam 表示垃圾短信（1），ham 表示非垃圾短信（0）
data_path = '../input/SMSSpamCollection'
label = '\t'
fo = FileOperate(data_path, label)
X, y = fo.load_data()
# 2、分割数据集，得到训练数据集与测试数据集
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=.25,
random_state=666)
# 开始训练
nb = NaiveBayes()
nb.fit(X_train, y_train)
# 开始预测
y_pred = nb.predict(X_test)
# 计算得分
score = accuracy_score(y_test, y_pred)
print('准确率：', score)
In [31]:
# 3.3 采用测试集验证模型离线指标
# 训练集AUC
probs_train= lr_model.predict_proba(X_train)
AUC1 = metrics.roc_auc_score(Y_train, probs_train[:,1])
print("Train Auc: %s"%(AUC1))
# 测试集AUC
probs_test= lr_model.predict_proba(X_test)
predict_test = lr_model.predict(X_test)
AUC2 = metrics.roc_auc_score(Y_test, probs_test[:,1])
print("Test Auc: %s"%(AUC2))
# 准确率
accuracy = metrics.accuracy_score(Y_test, predict_test)
print("Test Accuracy: %s"%(accuracy))
# 召回率
recall = metrics.recall_score(Y_test, predict_test)
print("Test Recall: %s"%(recall))
# F1值
f1 = metrics.f1_score(Y_test, predict_test)
print("Test F1: %s"%(f1))
In [42]:
# 3.4 打印模型参数
w=lr_model.coef_
print("参数大小:")
print(w.shape)
def score(self, X_test, y_test):
'''计算对测试集的预测准确度'''
y_predict = self.predict(X_test)
return accuracy_score(y_test, y_predict)
def score(self, X_test, y_test):
y_predict = self.predict(X_test)
return accuracy_score(y_test, y_predict)
lens, idx = torch.sort(lens, descending=True)
sents = sents[idx]
labs = labs[idx]
loss = model.neg_log_likelihood(sents, labs, lens)
loss.backward()
optimizer.step()
score, preds = model(sents, lens)
true_labs = [
seqid2text(labs[i, :l], ix_to_lab)
for i, l in enumerate(lens)
]
pred_labs = [
seqid2text(preds[i, :l], ix_to_lab)
for i, l in enumerate(lens)
]
acc = accuracy_score(true_labs, pred_labs)
f1 = f1_score(true_labs, pred_labs)
print(
"Epoch {}, batch {}, train loss {:.4f}, train acc {:.4f}, train f1 {:.4f} "
.format(epoch, i, loss.item(), acc, f1))
if ((i + 1) % 50 == 0):
with torch.no_grad():
model.eval()
print("Evaluation on validation set")
true_labels = []
pred_labels = []
for batch in val_data_loader:
sents, labs, lens = batch
sents = pad_sequence(sents,
batch_first=True).to(device)
labs = pad_sequence(labs,
#导入iris数据
iris = datasets.load_iris()
X = iris.data
y = iris.target
standardScaler = StandardScaler()
standardScaler.fit
#将数据集拆分为train、test数据
X_train, X_test, y_train, y_test = train_test_split(X, y, 0.2)
#数据的归一化（训练数据集、测试数据集的归一化）
standardScaler = StandardScaler()
standardScaler.fit(X_train)
X_train = standardScaler.transform(X_train)
X_test = standardScaler.transform(X_test)
print(X_test)
#使用上述数据进行训练和预测
kcf = kNNClassifier(3)
kcf.fit(X_train, y_train)
y_predict = kcf.predict(X_test)
#计算预测准确度
accuracy = accuracy_score(y_test, y_predict)
print('Origin: ', y_test)
print('Predict:', y_predict)
print('Accuracy:', accuracy)
print(kcf.score(X_test, y_test))
def score(self, X, y, sample_weight=None):
from metrics import accuracy_score
return accuracy_score(y, self.predict(X), sample_weight=sample_weight)
def score(self, X, y, sample_weight=None):
from metrics import accuracy_score
return accuracy_score(y, self.predict(X), sample_weight=sample_weight)
def score(self, x_test, y_test):
assert self._x_train is not None and self._y_train is not None, "must fit before score"
y_predict = self.predict(x_test)
return accuracy_score(y_test, y_predict)
data[:, :48],
columns=columns) # change nr of columns in file
y = data[:, 48] # last column in file
X_train, X_test, y_train, y_test = train_test_split(
df, y, test_size=0.2, random_state=0)
clf = RandomForestClassifier(n_estimators=ne,
criterion=crit,
max_features=mf,
min_samples_leaf=msl)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
results.append(
"criterion " + crit + ", n_estimators " + str(ne) +
", max_features " + mf + ", min_samples_leaf " + str(msl) +
", accuracy " +
str(metrics.accuracy_score(
y_test, y_pred, normalize=True)) + ", precision " +
str(metrics.precision_score(y_test, y_pred, pos_label="1"))
)
for res in results:
print(res)
#Decision Tree
"""
criterion = ['gini', 'entropy']
max_features = ['auto', 'sqrt', 'log2']
min_samples_leaf = [ 1, 2, 3, 4, 5]
for crit in criterion:
for mf in max_features:
for msl in min_samples_leaf:
def evaluate(args, model, eval_dataloader, params):
model.eval()
# 记录平均损失
loss_avg = utils.RunningAverage()
# init
pre_result = []
gold_result = []
# get data
for batch in tqdm(eval_dataloader, unit='Batch'):
# to device
batch = tuple(t.to(params.device) for t in batch)
input_ids, input_mask, segment_ids, start_pos, end_pos, ne_cate = batch
with torch.no_grad():
# get loss
loss = model(input_ids,
token_type_ids=segment_ids,
attention_mask=input_mask,
start_positions=start_pos,
end_positions=end_pos)
if params.n_gpu > 1 and args.multi_gpu:
loss = loss.mean() # mean() to average on multi-gpu.
# update the average loss
loss_avg.update(loss.item())
# inference
start_logits, end_logits = model(input_ids=input_ids,
token_type_ids=segment_ids,
attention_mask=input_mask)
# gold label
start_pos = start_pos.to("cpu").numpy().tolist()
end_pos = end_pos.to("cpu").numpy().tolist()
input_mask = input_mask.to('cpu').numpy().tolist()
ne_cate = ne_cate.to("cpu").numpy().tolist()
# predict label
start_label = start_logits.detach().cpu().numpy().tolist()
end_label = end_logits.detach().cpu().numpy().tolist()
# idx to label
cate_idx2label = {
idx: value
for idx, value in enumerate(params.label_list)
}
# get bio result
for start_p, end_p, start_g, end_g, input_mask_s, ne_cate_s in zip(
start_label, end_label, start_pos, end_pos, input_mask,
ne_cate):
ne_cate_str = cate_idx2label[ne_cate_s]
# 问题长度
q_len = len(IO2QUERY[ne_cate_str])
# 有效长度
act_len = sum(input_mask_s[q_len + 2:-1])
# get BIO labels
pre_bio_labels = pointer2bio(start_p[q_len + 2:q_len + 2 +
act_len],
end_p[q_len + 2:q_len + 2 + act_len],
ne_cate=ne_cate_str)
gold_bio_labels = pointer2bio(start_g[q_len + 2:q_len + 2 +
act_len],
end_g[q_len + 2:q_len + 2 + act_len],
ne_cate=ne_cate_str)
pre_result.append(pre_bio_labels)
gold_result.append(gold_bio_labels)
# metrics
f1 = f1_score(y_true=gold_result, y_pred=pre_result)
acc = accuracy_score(y_true=gold_result, y_pred=pre_result)
# f1, acc
metrics = {'loss': loss_avg(), 'f1': f1, 'acc': acc}
metrics_str = "; ".join("{}: {:05.2f}".format(k, v)
for k, v in metrics.items())
logging.info("- {} metrics: ".format('Val') + metrics_str)
# f1 classification report
report = classification_report(y_true=gold_result, y_pred=pre_result)
logging.info(report)
return metrics
def score(self, X_test, Y_test):
Y_predict = self.predict(X_test)
accuracy = accuracy_score(Y_test, Y_predict)
return accuracy
def score(self, X_test, y_test):
"""根据测试数据集 X_test 和 y_test 确定当前模型的准确度"""
y_predict = self.predict(X_test)
return accuracy_score(y_test, y_predict)
y_fold_train = y_train[train_fold_idx]
if kernel_method:
K_fold_test = K_train[test_fold_idx, :][:, train_fold_idx]
else:
X_fold_test = X_train[test_fold_idx]
y_fold_test = y_train[test_fold_idx]
if kernel_method:
kernel_model.fit(K_fold_train, y_fold_train)
y_pred = kernel_model.predict(K_fold_test)
else:
kernel_model.fit(X_fold_train, y_fold_train)
y_pred = kernel_model.predict(X_fold_test)
fold_score = metrics.accuracy_score(y_pred, y_fold_test)
print(fold_score)
cv_scores.append(fold_score)
cv_score_mean = round(np.mean(cv_scores), 5)
cv_score_std = round(np.std(cv_scores), 5)
cv_score_min = round(np.min(cv_scores), 5)
cv_score_max = round(np.max(cv_scores), 5)
print("\nMean score :", cv_score_mean)
print("Standard deviation :", cv_score_std)
print("Min score :", cv_score_min)
print("Max score :", cv_score_max)