In [118]:
import os
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID" # see issue #152
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
from tensorflow.python.client import device_lib
print(device_lib.list_local_devices())
In [1]:
from IPython.display import Image
def show(id, w=6):
return Image('../pics/class3/{}.jpg'.format(id), width=w*100)
In [2]:
!ls Recurrent_Neural_Networks/
In [3]:
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
In [4]:
dataset_train = pd.read_csv('Recurrent_Neural_Networks/Google_Stock_Price_Train.csv')
print dataset_train.shape
dataset_train.head()
Out[4]:
In [5]:
training_set = dataset_train.iloc[:,1:2].values
print training_set.shape, type(training_set)
training_set
Out[5]:
In [6]:
show(101)
Out[6]:
In [7]:
from sklearn.preprocessing import MinMaxScaler, Normalizer, StandardScaler
# MinMaxScaler?
In [8]:
sc = MinMaxScaler()
sc
Out[8]:
In [9]:
training_set_scaled = sc.fit_transform(training_set)
training_set_scaled
Out[9]:
In [10]:
training_set_scaled.shape
Out[10]:
In [11]:
# creating a data structure with 60 timesteps and 1 output
X_train = []
y_train = []
for i in range(60, training_set_scaled.shape[0]):
X_train.append(training_set_scaled[i-60:i, 0])
y_train.append(training_set_scaled[i, 0])
print len(X_train), len(y_train)
X_train, y_train = np.array(X_train), np.array(y_train)
X_train.shape, y_train.shape
Out[11]:
In [12]:
pd.DataFrame(X_train).tail()
Out[12]:
In [14]:
# reshaping
# np.reshape?
print np.reshape(X_train, (1198,60,1)).shape
X_train = np.reshape(X_train, (X_train.shape[0], X_train.shape[1], 1))
In [15]:
# build RNN
from keras.models import Sequential
from keras.layers import Dense, Dropout, LSTM
? return sequences
In [28]:
# initializing
regressor = Sequential()
# first LSTM layer
regressor.add(LSTM(units= 50, return_sequences= True, input_shape = (X_train.shape[1], 1)))
# add dropout
regressor.add(Dropout(0.2))
# second LSTM layer
regressor.add(LSTM(units= 50, return_sequences= True))
regressor.add(Dropout(0.2))
# third LSTM layer
regressor.add(LSTM(units= 50, return_sequences= True))
regressor.add(Dropout(0.2))
# fourth LSTM layer
# regressor.add(LSTM(units= 50, return_sequences= False))
regressor.add(LSTM(units= 50))
regressor.add(Dropout(0.2))
# output layer
regressor.add(Dense(units=1))
# compiling
regressor.compile(optimizer='adam', loss='mean_squared_error')
regressor.summary()
In [32]:
print X_train.shape, y_train.shape
regressor.fit(X_train, y_train, epochs= 100, batch_size= 32)
Out[32]:
In [33]:
# CPU is slower
# print X_train.shape, y_train.shape
# regressor.fit(X_train, y_train, epochs= 100, batch_size= 32)
In [37]:
regressor.save('RNN_stock.h5')
In [39]:
!ls -hl
In [121]:
from keras.models import load_model
In [188]:
# predict
dataset_test = pd.read_csv('Recurrent_Neural_Networks/Google_Stock_Price_Test.csv')
real_stopck_price= dataset_test.iloc[:,1:2].values
real_stopck_price.shape
Out[188]:
In [189]:
real_stopck_price[:3]
Out[189]:
In [126]:
loads = load_model('./RNN_stock.h5')
loads
Out[126]:
In [220]:
dataset_train = pd.read_csv('Recurrent_Neural_Networks/Google_Stock_Price_Train.csv')
print dataset_train.shape
dataset_train.head()
Out[220]:
In [225]:
train_open_last=dataset_train['Open'].values
train_open_last.shape
Out[225]:
In [226]:
train_open_last[-80:]==train_open_last[1258-20-60:]
Out[226]:
In [227]:
train_open_last = train_open_last[-80:].reshape(-1,1)
train_open_last.shape
Out[227]:
In [228]:
train_open_last.shape
Out[228]:
In [229]:
train_open_last[:3]
Out[229]:
In [230]:
train_open_last = sc.transform(train_open_last)
In [239]:
train_open_last[79]
Out[239]:
In [231]:
Y_test = []
for i in range(60, 80):
Y_test.append(train_open_last[i-60:i,0])
Y_test = np.array(Y_test)
Y_test.shape
Out[231]:
In [232]:
Y_test
Out[232]:
In [233]:
Y_test=Y_test.reshape(20,60,1)
Y_test.shape
Out[233]:
In [234]:
pred = loads.predict(Y_test)
pred.shape
Out[234]:
In [235]:
pred[:3]
Out[235]:
In [236]:
pred = sc.inverse_transform(pred)
pred[:3]
Out[236]:
In [256]:
plt.figure(figsize=(10,4))
plt.plot(pred, c='r', label='pred')
plt.plot(real_stopck_price, c='g', label='read')
plt.legend(loc=2)
plt.title('prediction')
plt.show()
In [251]:
plt.legend?