def svm_predict(days=10, stock='YHOO'):
# Generate sample data
lag = 2
offset = 20
closingPrices, stockFeature = stockFeatures(stock, lag)
# set training dat size
N = len(stockFeature) / 3 * 2
# training data
y = closingPrices[lag:(N + lag)]
X = stockFeature[0:N]
# test data
Z = stockFeature[N + offset:(N + days + offset)]
# Fit regression model
svr_rbf = SVR(kernel='rbf', C=1e2, gamma=0.018)
y_rbf = svr_rbf.fit(X, y).predict(Z)
a = [round(i, 3) for i in list(y_rbf)]
ax = plt.subplot(111)
pre = 10
b1 = closingPrices[N + lag + offset - pre:N + lag + offset]
b2 = closingPrices[N + lag + offset - pre:N + lag + offset]
b1.extend(a)
b2.extend(closingPrices[N + lag + offset:N + lag + days + offset])
plt.plot(b1, label='Prediction ' + stock + " " + str(lag))
plt.plot(b2, label='Origin ' + stock + ' ' + str(lag))
plt.legend()
plt.show()
return a
def mlp_predict(days, stock='YHOO'):
# Generate sample data
lag = 5
offset = 20
closingPrices, stockFeature = stockFeatures(stock, lag)
#set training dat size
N = len(stockFeature) / 3 * 2
#training data
y = closingPrices[lag:(N + lag)]
X = stockFeature[0:N]
#test data
Z = stockFeature[N + offset:(N + days + offset)]
# Fit regression model hidden_layer_sizes
mlp = MLPRegressor(solver='lbfgs',
activation='identity',
verbose=False,
max_iter='tol',
hidden_layer_sizes=(50, ))
mlp.fit(X, y)
y_log = mlp.predict(Z)
#print np.ndarray.tolist(y_log)
a = [round(i, 3) for i in list(y_log)]
ax = plt.subplot(111)
pre = 10
b1 = closingPrices[N + lag + offset - pre:N + lag + offset]
b2 = closingPrices[N + lag + offset - pre:N + lag + offset]
b1.extend(a)
b2.extend(closingPrices[N + lag + offset:N + lag + days + offset])
plt.plot(b1, label='Prediction ' + stock + " " + str(lag))
plt.plot(b2, label='Origin ' + stock + ' ' + str(lag))
plt.legend()
plt.show()
return a
def svm_predict(days=10, stock='YHOO'):
# Generate sample data
lag = 10
closingPrices, stockFeature = stockFeatures(stock, lag)
N = len(stockFeature) / 2
y = closingPrices[lag:(N + lag)]
X = stockFeature[0:N]
Z = stockFeature[N:(N + days)]
# Fit regression model
svr_rbf = SVR(kernel='rbf', C=1e2, gamma=0.018)
y_rbf = svr_rbf.fit(X, y).predict(Z)
a = [round(i, 3) for i in list(y_rbf)]
ax = plt.subplot(111)
plt.plot(a, label='Prediction ' + stock + " " + str(lag))
plt.plot(closingPrices[N + lag:N + lag + days],
label='Origin ' + stock + ' ' + str(lag))
plt.legend()
plt.show()
return a
def mlp_predict(days, stock='YHOO'):
# Generate sample data
lag = 5
offset = 0
closingPrices, stockFeature = stockFeatures(stock, lag)
N = len(stockFeature) / 3 * 2
upDown = []
for i in range(1, len(closingPrices)):
upDown.append(-1 if (closingPrices[i - 1] > closingPrices[i]) else 0)
trainLable = upDown[lag - 1:(N + lag - 1)]
trainData = stockFeature[0:N]
testData = stockFeature[N + offset:(N + days + offset)]
testLable = upDown[N + lag + offset - 1:N + lag + days + offset - 1]
# Fit model
mlp = MLPClassifier(solver='lbfgs',
activation='relu',
verbose=False,
max_iter='tol',
hidden_layer_sizes=())
mlp.fit(trainData, trainLable)
y_pre = mlp.predict(testData).tolist()
print mlp.score(testData, testLable)
return y_pre