def on_epoch_end(self, epoch, logs={}):
"""This function will be called after each epoch.
:param epoch: the index of the epoch
:type epoch: int
:param logs: contain keys for quantities relevant to the current batch or epoch
:type logs: dict
:return: the model file will be updated if necessary
:rtype: None
"""
model_path = self.model_path.format(epoch=epoch, **logs)
probability_estimates = self.model.predict_proba(self.X_test,
verbose=0)
prediction = probability_estimates[:, 1]
score = evaluation.compute_Weighted_AUC(self.Y_test, prediction)
if self.best_score < 0 or score > self.best_score:
print("In epoch {:05d}: {} improved from {:.4f} to {:.4f}, saving model to {}.".format(\
epoch + 1, self.monitor, self.best_score, score, os.path.basename(model_path)))
self.best_score_index = epoch + 1
self.best_score = score
self.model.save_weights(model_path, overwrite=True)
else:
pass
def on_epoch_end(self, epoch, logs={}):
"""This function will be called after each epoch.
:param epoch: the index of the epoch
:type epoch: int
:param logs: contain keys for quantities relevant to the current batch or epoch
:type logs: dict
:return: the model file will be updated if necessary
:rtype: None
"""
model_path = self.model_path.format(epoch=epoch, **logs)
probability_estimates = self.model.predict_proba(self.X_test, verbose=0)
prediction = probability_estimates[:, 1]
score = evaluation.compute_Weighted_AUC(self.Y_test, prediction)
if self.best_score < 0 or score > self.best_score:
print(
"In epoch {:05d}: {} improved from {:.4f} to {:.4f}, saving model to {}.".format(
epoch + 1, self.monitor, self.best_score, score, os.path.basename(model_path)
)
)
self.best_score_index = epoch + 1
self.best_score = score
self.model.save_weights(model_path, overwrite=True)
else:
pass
def train_model(X_train, Y_train, X_test, Y_test, model_path):
"""Training phase.
:param X_train: the training attributes
:type X_train: numpy array
:param Y_train: the training labels
:type Y_train: numpy array
:param X_test: the testing attributes
:type X_test: numpy array
:param Y_test: the testing labels
:type Y_test: numpy array
:param model_path: the path of the model file
:type model_path: string
:return: best_score refers to the highest score
:rtype: float
"""
# Set the parameters for SVC
param_grid = [{"C": [1, 10, 100, 1000], "gamma": ["auto"], "kernel": ["linear"]}, \
{"C": [1, 10, 100, 1000], "gamma": [0.001, 0.0001], "kernel": ["rbf"]}]
parameters_combinations = ParameterGrid(param_grid)
# Get a list of different classifiers
unique_classifier_list = []
for current_parameters in parameters_combinations:
current_classifier = SVC(C=current_parameters["C"],
kernel=current_parameters["kernel"],
gamma=current_parameters["gamma"],
probability=True)
unique_classifier_list.append(current_classifier)
# Loop through the classifiers
best_score = -np.Inf
for classifier_index, classifier in enumerate(unique_classifier_list):
classifier.fit(X_train, Y_train)
probability_estimates = classifier.predict_proba(X_test)
prediction = probability_estimates[:, 1]
score = evaluation.compute_Weighted_AUC(Y_test, prediction)
print("Classifier {:d} achieved {:.4f}.".format(
classifier_index, score))
if best_score < 0 or score > best_score:
print("Score improved from {:.4f} to {:.4f}, saving model to {}.".format(\
best_score, score, os.path.basename(model_path)))
best_score = score
joblib.dump(classifier, model_path)
return best_score
def train_model(X_train, Y_train, X_test, Y_test, model_path):
"""Training phase.
:param X_train: the training attributes
:type X_train: numpy array
:param Y_train: the training labels
:type Y_train: numpy array
:param X_test: the testing attributes
:type X_test: numpy array
:param Y_test: the testing labels
:type Y_test: numpy array
:param model_path: the path of the model file
:type model_path: string
:return: best_score refers to the highest score
:rtype: float
"""
# Set the parameters for SVC
param_grid = [{"C": [1, 10, 100, 1000], "gamma": ["auto"], "kernel": ["linear"]}, \
{"C": [1, 10, 100, 1000], "gamma": [0.001, 0.0001], "kernel": ["rbf"]}]
parameters_combinations = ParameterGrid(param_grid)
# Get a list of different classifiers
unique_classifier_list = []
for current_parameters in parameters_combinations:
current_classifier = SVC(C=current_parameters["C"],
kernel=current_parameters["kernel"],
gamma=current_parameters["gamma"],
probability=True)
unique_classifier_list.append(current_classifier)
# Loop through the classifiers
best_score = -np.Inf
for classifier_index, classifier in enumerate(unique_classifier_list):
classifier.fit(X_train, Y_train)
probability_estimates = classifier.predict_proba(X_test)
prediction = probability_estimates[:, 1]
score = evaluation.compute_Weighted_AUC(Y_test, prediction)
print("Classifier {:d} achieved {:.4f}.".format(classifier_index, score))
if best_score < 0 or score > best_score:
print("Score improved from {:.4f} to {:.4f}, saving model to {}.".format(\
best_score, score, os.path.basename(model_path)))
best_score = score
joblib.dump(classifier, model_path)
return best_score
