def main():
# create feature extractor for Google
extractor = fe.featureExtractor(0)
# get features, label and corresponding date
x, y, date = extractor.getFeature(1, 1)
ts = tester.Tester(5)
ts.test(regressionModel(0, gradient_type=0, zscore=True), x, y, 0)
def main():
# create feature extractor for Google
extractor = fe.featureExtractor(0)
# get features, label and corresponding date
x, y, date = extractor.getFeature(1, 1)
# y = tools.sign(y)
ts = tester.Tester(5)
ts.test(MLNN(hidden_layers=(2, 2), zscore=False), x, y, 0)
def main():
# create feature extractor for Google
extractor = fe.featureExtractor(0)
# get features, label and corresponding date
x, y, date = extractor.getFeature(1, 1)
y = tools.sign(y)
ts = tester.Tester(5)
ts.test(SVM(1, zscore=True), x, y, 1)
def testSingleModel(self, fluc=5, sentiment=1, data_set_size=0):
sentimentals = {
0: 'positive/negative',
1: 'multiple moods',
2: '2d term frequency',
3: '4d term frequency',
}
print("-" * 60)
print("{}-day fluctuation, sentimental analysis mathod: {}".format(
fluc, sentimentals[sentiment]))
# linear regression model
model = TensorFlowNN(0.01, tf.tanh, 2, 2)
for i in range(0, len(TEST_COMPANY)):
# create feature extractor for current company
extractor = fe.featureExtractor(i)
# get features, label and corresponding date
x, y, date = extractor.getFeature(fluc, sentiment)
# set dataset size
data_set_size = min(data_set_size,
x.shape[0]) if data_set_size else x.shape[0]
x = x[:data_set_size]
y = y[:data_set_size]
# Test TensorFlow Neural Network
print("# TensorFlow Neural Network Tester with " + TEST_COMPANY[i])
model.train(x, y)
# predicted_y: [y1, y2, y3, ..., yn]
predicted_y = model.predict(x)
# plot the result
msg = "Fluctuation: " + str(fluc) + "-day, Sentiment: " + \
str(sentiment) + ", Dataset: " + str(data_set_size)
# tools.plot(predicted_y, y.values.ravel(), msg)
# computeMSE need input params in excatly same dimension
# Here both predicted_y and test_y.values.ravel() is a
# n-dimension vector [y1, y2, ..., yn]
mse = tools.computeMSE(predicted_y, y.values.ravel())
print("MSE:", mse)
def testTensorFlow(self,
stepSize,
hiddenLayers,
layerNodes,
activation_function,
fluc=5,
sentiment=0):
print('########## tesing tensorflow regression ##########')
print('stepSize: ' + str(stepSize))
print('hiddenLayers: ' + str(hiddenLayers))
print('layerNodes: ' + str(layerNodes))
print('activation_function: ' + str(activation_function))
nn = TensorFlowNN(stepSize, activation_function, hiddenLayers,
layerNodes)
for i in range(0, len(TEST_COMPANY)):
print("# TensorFlowNN Tester with " + TEST_COMPANY[i])
extractor = fe.featureExtractor(i)
x, y, date = extractor.getFeature(fluc, sentiment)
self.test(nn, x, y, 0)
def testAllModels(self, fluc=5, sentiment=1, data_set_size=0):
# linear regression model
model_lreg = reg.regressionModel(0, gradient_type=0, zscore=True)
model_svm = svm.SVM(1, zscore=True)
model_nn = mlnn.MLNN(hidden_layers=(5, ), zscore=True)
model_tfnn = TensorFlowNN(0.01, tf.tanh, 2, 2)
for i in range(0, len(TEST_COMPANY)):
# create feature extractor for current company
extractor = fe.featureExtractor(i)
# get features, label and corresponding date
x, y, date = extractor.getFeature(fluc, sentiment)
# set dataset size
data_set_size = min(data_set_size,
x.shape[0]) if data_set_size else x.shape[0]
x = x[:data_set_size]
y = y[:data_set_size]
# convert numerical label into discrete value for classifier
discrete_y = tools.sign(y)
# Test Linear Regression
print("# Linear Regression Tester with " + TEST_COMPANY[i])
self.test(model_lreg, x, y, 0, "Linear Regression")
print("-" * 60)
# Test TensorFlow Neural Network
print("# TensorFlow Neural Network Tester with " + TEST_COMPANY[i])
self.test(model_tfnn, x, y, 0, "TensorFlow Neural Network")
print("-" * 60)
# Test SVM
print("# SVM Tester with " + TEST_COMPANY[i])
self.test(model_svm, x, discrete_y, 1)
print("-" * 60)
# Test multi-layer nueral network
print("# Neural Network Tester with " + TEST_COMPANY[i])
self.test(model_nn, x, discrete_y, 1)
print("-" * 60)