-- coding: utf-8 --
@Date: 2019-01-18 14:17:56
@last Modified by: boyac
@last Modified time: 2019-02-21 17:38:57
from FinMind.Data import Load
import numpy as np
from keras.layers.core import Dense, Dropout
from keras.layers import LSTM , GRU
from keras.models import Sequential
from sklearn.preprocessing import MinMaxScaler
import matplotlib.pyplot as plt
import math
from keras.optimizers import Adam
print('input data')
data = Load.FinData(
dataset = 'TaiwanStockPrice',
select = ['2330.TW'],
date = '2010-01-01')
#colname = ['date', 'open', 'high', 'low', 'close', 'volume']
#data = data[colname]
print('select close price')
stock_price = data['Close'].values.astype('float32')
stock_price = stock_price.reshape(len(stock_price), 1)
print(' 畫圖 ')
plt.plot(stock_price)
plt.show()
print(' 歸一化 ')
scaler = MinMaxScaler(feature_range=(0, 1))
stock_price = scaler.fit_transform(stock_price)
print(' 取 80% data 當作 training data, 20% data 當作 testing data 做模型驗證 ')
train_size = int(len(stock_price) * 0.8)
test_size = len(stock_price) - train_size
train, test = stock_price[0:train_size,:], stock_price[train_size:len(stock_price),:]
print('切 data，拿前五天的股價，預測未來1天的股價，先做個 demo')
def process_data(data , n_features,future_days):
dataX, dataY = [], []
for i in range(len(data)-n_features-future_days):
a = data[i:(i+n_features), 0]
dataX.append(a)
dataY.append(data[i + n_features-1+future_days, 0])
return np.array(dataX), np.array(dataY)
reshape into X=t and Y=t+5
n_features = 5
future_days = 1
trainX, trainY = process_data(train, n_features,future_days)
testX, testY = process_data(test, n_features,future_days)
print(trainX.shape , trainY.shape , testX.shape , testY.shape)
lstm need input to have 3 dimensions
print('轉換成 LSTM 建模所需 data 的型態')
trainX = np.reshape(trainX, (trainX.shape[0], 1, trainX.shape[1]))
testX = np.reshape(testX, (testX.shape[0], 1, testX.shape[1]))
print('設定參數')
filepath="stock_weights.hdf5"
from keras.callbacks import ReduceLROnPlateau , ModelCheckpoint
lr_reduce = ReduceLROnPlateau(monitor='val_loss', factor=0.1, epsilon=0.0001, patience=1, verbose=1)
checkpoint = ModelCheckpoint(filepath, monitor='val_loss', verbose=1, save_best_only=True, mode='max')
print('建立 DL 模型，使用 RNN 常見的 GRU and LSTM')
model = Sequential()
model.add(GRU(256 , input_shape = (1 , n_features), return_sequences=True))
model.add(Dropout(0.2))
model.add(LSTM(2048))
model.add(Dropout(0.2))
model.add(Dense(64 , activation = 'relu'))
model.add(Dense(1))
#model.load_weights('/home/linsam/job/stock_weight.h5')
model.compile(loss='mean_squared_error', optimizer=Adam(lr = 0.0005) , metrics = ['mean_squared_error'])
print('training')
history = model.fit(trainX, trainY, epochs=5 , batch_size = 128 ,
callbacks = [checkpoint , lr_reduce] , validation_data = (testX,testY))
print(' save weight ')
model.save_weights('stock_weight.h5')
print('計算 error')
def model_score(model, X_train, y_train, X_test, y_test):
trainScore = model.evaluate(X_train, y_train, verbose=0)
print('Train Score: %.5f MSE (%.2f RMSE)' % (trainScore[0], math.sqrt(trainScore[0])))
testScore = model.evaluate(X_test, y_test, verbose=0)
print('Test Score: %.5f MSE (%.2f RMSE)' % (testScore[0], math.sqrt(testScore[0])))
return trainScore[0], testScore[0]
tem = model_score(model, trainX, trainY , testX, testY)
pred = model.predict(testX)
pred = scaler.inverse_transform(pred)
print('\n Actual Stock Prices')
print( pred[:10] )
testY2 = testY.reshape(testY.shape[0] , 1)
testY2 = scaler.inverse_transform(testY2)
print('\n Predicted Stock Prices ')
print( testY2[:10] )
print('畫圖')
print("\nRed - Predicted Stock Prices, Blue - Actual Stock Prices")
plt.rcParams["figure.figsize"] = (15,7)
plt.plot(testY2 , 'b')
plt.plot(pred , 'r')
plt.xlabel('Time')
plt.ylabel('Stock Prices')
plt.title('Red - Predicted Stock Prices, Blue - Actual Stock Prices')
plt.grid(True)
plt.show()
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent