def train_test_score(classifier, data, target, labels, clf2, ml_seed):
normalizer = make_normalizer(data)
split_point = round(ml_seed.split_ratio * len(data))
train_set = data[:split_point]
train_key = target[:split_point]
test_set = data[split_point:]
test_key = target[split_point:]
if ml_seed.normalize_my_data:
my_train_set = normalizer(train_set)
my_test_set = normalizer(test_set)
else:
my_train_set = train_set
my_test_set = test_set
classifier.train(my_train_set, train_key, ml_seed.relation)
prediction = classifier.predict(my_test_set)
if ml_seed.normalize_their_data:
their_train_set = normalizer(train_set)
their_test_set = normalizer(test_set)
else:
their_train_set = train_set
their_test_set = test_set
clf2.fit(their_train_set, train_key)
prediction_real = clf2.predict(their_test_set)
accuracy = calc_accuracy(test_key, prediction)
accuracy_real = calc_accuracy(test_key, prediction_real)
confusion_matrix = make_confusion_matrix(labels, test_key, prediction)
confusion_matrix_real = make_confusion_matrix(labels, test_key,
prediction_real)
return accuracy, accuracy_real, confusion_matrix, confusion_matrix_real
def train_test_score(classifier, data, target, labels, clf2, ml_seed):
normalizer = make_normalizer(data)
split_point = round(ml_seed.split_ratio * len(data))
train_set = data[:split_point]
train_key = target[:split_point]
test_set = data[split_point:]
test_key = target[split_point:]
if ml_seed.normalize_my_data:
my_train_set = normalizer(train_set)
my_test_set = normalizer(test_set)
else:
my_train_set = train_set
my_test_set = test_set
classifier.train(my_train_set, train_key, ml_seed.relation)
prediction = classifier.predict(my_test_set)
if ml_seed.normalize_their_data:
their_train_set = normalizer(train_set)
their_test_set = normalizer(test_set)
else:
their_train_set = train_set
their_test_set = test_set
clf2.fit(their_train_set, train_key)
prediction_real = clf2.predict(their_test_set)
accuracy = calc_accuracy(test_key, prediction)
accuracy_real = calc_accuracy(test_key, prediction_real)
confusion_matrix = make_confusion_matrix(labels, test_key, prediction)
confusion_matrix_real = make_confusion_matrix(
labels, test_key, prediction_real)
return accuracy, accuracy_real, confusion_matrix, confusion_matrix_real
def cross_val_score(classifier, data, target, labels, clf2, ml_seed):
k = max(0, min(ml_seed.cross_folds, len(data)))
step = len(data) // k
results = np.zeros(k)
results_real = np.zeros(k)
confusion_matrices = []
confusion_matrices_real = []
normalizer = make_normalizer(data)
for i, pos in zip(range(0, len(data), step), range(k)):
train_set = np.vstack((data[:i], data[i + step:]))
train_key = np.hstack((target[:i], target[i + step:]))
test_set = data[i:i + step]
test_key = target[i:i + step]
if ml_seed.normalize_my_data:
my_train_set = normalizer(train_set)
my_test_set = normalizer(test_set)
else:
my_train_set = train_set
my_test_set = test_set
classifier.train(my_train_set, train_key, ml_seed.relation)
prediction = classifier.predict(my_test_set)
results[pos] = calc_accuracy(test_key, prediction)
if ml_seed.normalize_their_data:
their_train_set = normalizer(train_set)
their_test_set = normalizer(test_set)
else:
their_train_set = train_set
their_test_set = test_set
clf2.fit(their_train_set, train_key)
prediction_real = clf2.predict(their_test_set)
results_real[pos] = calc_accuracy(test_key, prediction_real)
confusion_matrices.append(
make_confusion_matrix(labels, test_key, prediction))
confusion_matrices_real.append(
make_confusion_matrix(labels, test_key, prediction_real))
confusion_matrix = flatten_confusion_matrices(confusion_matrices)
confusion_matrix_real = flatten_confusion_matrices(confusion_matrices_real)
if ml_seed.confusion_matrix_format != CMF_ACTUAL:
confusion_matrix.values = convert_to_zero_one(confusion_matrix.values)
confusion_matrix_real.values = convert_to_zero_one(
confusion_matrix_real.values)
return results, results_real, confusion_matrix, confusion_matrix_real
def cross_val_score(classifier, data, target, labels, clf2, ml_seed):
k = max(0, min(ml_seed.cross_folds, len(data)))
step = len(data) // k
results = np.zeros(k)
results_real = np.zeros(k)
confusion_matrices = []
confusion_matrices_real = []
normalizer = make_normalizer(data)
for i, pos in zip(range(0, len(data), step), range(k)):
train_set = np.vstack((data[:i], data[i+step:]))
train_key = np.hstack((target[:i], target[i+step:]))
test_set = data[i:i+step]
test_key = target[i:i+step]
if ml_seed.normalize_my_data:
my_train_set = normalizer(train_set)
my_test_set = normalizer(test_set)
else:
my_train_set = train_set
my_test_set = test_set
classifier.train(my_train_set, train_key, ml_seed.relation)
prediction = classifier.predict(my_test_set)
results[pos] = calc_accuracy(test_key, prediction)
if ml_seed.normalize_their_data:
their_train_set = normalizer(train_set)
their_test_set = normalizer(test_set)
else:
their_train_set = train_set
their_test_set = test_set
clf2.fit(their_train_set, train_key)
prediction_real = clf2.predict(their_test_set)
results_real[pos] = calc_accuracy(test_key, prediction_real)
confusion_matrices.append(
make_confusion_matrix(labels, test_key, prediction))
confusion_matrices_real.append(
make_confusion_matrix(labels, test_key, prediction_real))
confusion_matrix = flatten_confusion_matrices(confusion_matrices)
confusion_matrix_real = flatten_confusion_matrices(confusion_matrices_real)
if ml_seed.confusion_matrix_format != CMF_ACTUAL:
confusion_matrix.values = convert_to_zero_one(confusion_matrix.values)
confusion_matrix_real.values = convert_to_zero_one(
confusion_matrix_real.values)
return results, results_real, confusion_matrix, confusion_matrix_real
