def main():
X, y = get_small_data('TIMESNOW')
n_features = list()
for i in range(1,21):
n_features.append(i*10)
n_features.append(228)
clf = RandomForestClassifier(n_estimators=10)
t0 = time.clock()
result = do_PCA_and_cv(clf, X.toarray(), y, n_features)
testTime = time.clock() - t0
print('Total time: ' + str(testTime))
scores = result[0]
scores_std = result[1]
plt.clf()
plt.plot(n_features, scores)
plt.plot(n_features, np.array(scores) + np.array(scores_std), 'b--')
plt.plot(n_features, np.array(scores) - np.array(scores_std), 'b--')
locs, labels = plt.yticks()
plt.yticks(locs, list(map(lambda x: "%g" % x, locs)))
plt.ylabel('CV score')
plt.xlabel('Parameter N')
plt.show()
def main():
X, y = get_small_data('TIMESNOW')
n_features = list()
for i in range(1, 21):
n_features.append(i * 10)
n_features.append(228)
clf = RandomForestClassifier(n_estimators=10)
t0 = time.clock()
result = do_PCA_and_cv(clf, X.toarray(), y, n_features)
testTime = time.clock() - t0
print('Total time: ' + str(testTime))
scores = result[0]
scores_std = result[1]
plt.clf()
plt.plot(n_features, scores)
plt.plot(n_features, np.array(scores) + np.array(scores_std), 'b--')
plt.plot(n_features, np.array(scores) - np.array(scores_std), 'b--')
locs, labels = plt.yticks()
plt.yticks(locs, list(map(lambda x: "%g" % x, locs)))
plt.ylabel('CV score')
plt.xlabel('Parameter N')
plt.show()
def test_method(clf, name):
channels = ( 'NDTV', 'TIMESNOW', 'CNNIBN', 'CNN', 'BBC' )
for channel in channels:
print('\nProcessing channel {} with method {}'.format(channel, name))
#f.write('\nProcessing channel {}\n'.format(channel))
X, y = get_small_data(channel)
n_features = list()
for i in range(1,21):
n_features.append(i*10)
n_features.append(X.shape[1])
selector = RandomForestClassifier(n_estimators=60)
selector.fit(X, y)
t0 = time.clock()
result = do_randfor_selection(clf, X.toarray(), y, n_features, selector.feature_importances_)
testTime = time.clock() - t0
print('Total time: ' + str(testTime))
scores = result[0]
scores_std = result[1]
times = result[2]
f = open('RFS-{}-{}.log'.format(name, channel), 'at')
#f.write('N_features; mean_score; std error; mean training time\n')
for i in range(0, len(n_features)):
f.write('{}; {}; {}; {}\n'.format(n_features[i], scores[i], scores_std[i], times[i]))
f.close()
plt.clf()
plt.plot(n_features, scores)
plt.plot(n_features, np.array(scores) + np.array(scores_std), 'b--')
plt.plot(n_features, np.array(scores) - np.array(scores_std), 'b--')
locs, labels = plt.yticks()
plt.yticks(locs, list(map(lambda x: "%g" % x, locs)))
plt.ylabel('CV score')
plt.xlabel('Parameter N')
#plt.show()
plt.savefig('{}_{}_r_selection.png'.format(name, channel))
plt.clf()
plt.plot(n_features, times)
locs, labels = plt.yticks()
plt.yticks(locs, list(map(lambda x: "%g" % x, locs)))
plt.ylabel('Mean learning time')
plt.xlabel('Parameter N')
plt.savefig('{}_{}_r_selection_time.png'.format(name, channel))
def main():
if len(sys.argv) < 3:
print('Invalid params. USAGE: param_optimization.py <method> <channel>')
print('Available methods:\n\tknn, lda, svm, randfor, gtb')
print('Available channels:\n\tcnn, bbc, ndtv, timesnow, cnnibn')
sys.exit()
method_name, channel_name = sys.argv[1], sys.argv[2]
optimizers = { 'knn': KNNOptimizer, 'lda': LDAOptimizer, 'svm': SVMOptimizer,
'randfor': RandForestOptimizer, 'gtb': GTBOptimizer }
X, y = get_small_data(channel_name.upper())
optimizer = optimizers[method_name]()
optimizer.optimize(X, y)
optimizer.log_results(channel_name)
def test_method(clf, name):
channels = ('NDTV', 'TIMESNOW', 'CNNIBN', 'CNN', 'BBC')
f = open('PCA_results_{}.txt'.format(name), 'w')
for channel in channels:
print('\nProcessing channel {}'.format(channel))
f.write('\nProcessing channel {}\n'.format(channel))
X, y = get_small_data(channel)
n_features = list()
for i in range(1, 21):
n_features.append(i * 10)
n_features.append(227)
t0 = time.clock()
result = do_PCA_and_cv(clf, X.toarray(), y, n_features)
testTime = time.clock() - t0
print('Total time: ' + str(testTime))
scores = result[0]
scores_std = result[1]
times = result[2]
f.write('N_features; mean_score; std error; mean training time\n')
for i in range(0, len(n_features)):
f.write('{}; {}; {}; {}\n'.format(n_features[i], scores[i],
scores_std[i], times[i]))
plt.clf()
plt.plot(n_features, scores)
plt.plot(n_features, np.array(scores) + np.array(scores_std), 'b--')
plt.plot(n_features, np.array(scores) - np.array(scores_std), 'b--')
locs, labels = plt.yticks()
plt.yticks(locs, list(map(lambda x: "%g" % x, locs)))
plt.ylabel('CV score')
plt.xlabel('Parameter N')
#plt.show()
plt.savefig('{}_{}_PCA.png'.format(name, channel))
plt.clf()
plt.plot(n_features, times)
locs, labels = plt.yticks()
plt.yticks(locs, list(map(lambda x: "%g" % x, locs)))
plt.ylabel('Mean learning time')
plt.xlabel('Parameter N')
plt.savefig('{}_{}_PCA_time.png'.format(name, channel))
f.close()
def main():
if len(sys.argv) < 3:
print(
'Invalid params. USAGE: param_optimization.py <method> <channel>')
print('Available methods:\n\tknn, lda, svm, randfor, gtb')
print('Available channels:\n\tcnn, bbc, ndtv, timesnow, cnnibn')
sys.exit()
method_name, channel_name = sys.argv[1], sys.argv[2]
optimizers = {
'knn': KNNOptimizer,
'lda': LDAOptimizer,
'svm': SVMOptimizer,
'randfor': RandForestOptimizer,
'gtb': GTBOptimizer
}
X, y = get_small_data(channel_name.upper())
optimizer = optimizers[method_name]()
optimizer.optimize(X, y)
optimizer.log_results(channel_name)
