def once_a_hour():
log.info("main", "once_a_hour")
logline = ""
for i in range(20):
logline += str(values[i]) + ";"
logline = logline.replace(' ', '')
log.line(logline)
return
def load_data(feature, classes, components_cnt):
log.line()
log.dataset_configuration(feature, classes, components_cnt)
log.message('Loading data...')
start = time.time()
train_data, train_labels, test_data, test_labels = d.read_test_and_train_data(feature, classes)
log.time(time.time() - start)
log.dataset_summary(len(train_data), len(test_data))
return train_data, train_labels, test_data, test_labels
def transform_data(data, components_cnt):
train_data, train_labels, test_data, test_labels = data
log.line()
start = time.time()
log.message('pca...')
train_data, test_data = transform.perform_pca(train_data, test_data, components_cnt)
log.time(time.time() - start)
start = time.time()
log.message('normalize...')
train_data, test_data = transform.normalize_data(train_data, test_data)
log.time(time.time() - start)
log.dataset_summary(len(train_data), len(test_data))
return train_data, train_labels, test_data, test_labels
def transform_data_kernel(data, components_cnt, kernel_params):
train_data, train_labels, test_data, test_labels = data
log.line()
start = time.time()
log.message('pca...')
if kernel_params[0] == 'chi2':
train_data, test_data = transform.normalize_data(train_data, test_data)
train_data, test_data = transform.perform_kernel_pca(train_data, test_data, components_cnt, kernel_params)
log.time(time.time() - start)
start = time.time()
log.message('normalize...')
train_data, test_data = transform.normalize_data(train_data, test_data)
log.time(time.time() - start)
log.dataset_summary(len(train_data), len(test_data))
return train_data, train_labels, test_data, test_labels
def parameters_search(score, params, data, pair, n_jobs=8):
# Create a pipeline where our custom predefined kernel Chi2Kernel
# is run before SVC.
pipe = Pipeline([
('chi2', Chi2Kernel()),
('svm', svm.SVC()),
])
A_train, y_train, A_test, y_test = data
log.line()
print("# Tuning hyper-parameters for %s\n" % score)
clf = GridSearchCV(pipe, params, cv=5, scoring=score, n_jobs=n_jobs)
start = time.time()
clf.fit(A_train, y_train)
log.time(time.time() - start)
print("Best parameters set found on development set:\n%s\n" %
clf.best_params_)
print("\nGrid scores on development set:\n")
means = clf.cv_results_['mean_test_score']
stds = clf.cv_results_['std_test_score']
for mean, std, params in zip(means, stds, clf.cv_results_['params']):
print("%0.3f (+/-%0.03f) for %r" % (mean, std * 2, params))
print()
y_pred = clf.predict(A_test)
report = classification_report(y_test, y_pred)
print(report)
with open(const.RESULTS_FILE_NAME, "a") as file:
file.write(str(pair) + "\n")
for mean, std, params in zip(means, stds, clf.cv_results_['params']):
file.write("%0.3f (+/-%0.03f) for %r\n" % (mean, std * 2, params))
file.write("\n%s\n\n" % clf.best_params_)
file.write(report + "\n\n")
return clf
def parameters_search(score, params, data, pair, n_jobs=8):
train_data, train_labels, test_data, test_labels = data
log.line()
print("# Tuning hyper-parameters for %s\n" % score)
clf = GridSearchCV(
svm.SVC(C=1),
params,
cv=5,
scoring=score,
n_jobs=n_jobs
)
start = time.time()
clf.fit(train_data, train_labels)
log.time(time.time() - start)
print("Best parameters set found on development set:\n%s\n" % clf.best_params_)
print("\nGrid scores on development set:\n")
means = clf.cv_results_['mean_test_score']
stds = clf.cv_results_['std_test_score']
for mean, std, params in zip(means, stds, clf.cv_results_['params']):
print("%0.3f (+/-%0.03f) for %r" % (mean, std * 2, params))
pred_labels = clf.predict(test_data)
report = classification_report(test_labels, pred_labels)
print(report)
with open(const.RESULTS_FILE_NAME, "a") as file:
file.write(str(pair) + "\n")
for mean, std, params in zip(means, stds, clf.cv_results_['params']):
file.write("%0.3f (+/-%0.03f) for %r\n" % (mean, std * 2, params))
file.write("\n%s\n\n" % clf.best_params_)
file.write(report + "\n\n")
return clf