【转】基于keras 的神经网络股价预测模型

 from matplotlib.dates import DateFormatter, WeekdayLocator, DayLocator, MONDAY,YEARLY
 from matplotlib.finance import quotes_historical_yahoo_ohlc, candlestick_ohlc
 #import matplotlib
 import tushare as ts
 import pandas as pd
 import matplotlib.pyplot as plt
 from matplotlib.pylab import date2num
 import datetime
 import numpy as np
 from pandas import DataFrame
 from numpy import row_stack,column_stack

 df=ts.get_hist_data('',start='2016-06-15',end='2017-11-06')
 dd=df[['open','high','low','close']]

 #print(dd.values.shape[0])

 dd1=dd .sort_index()

 dd2=dd1.values.flatten()

 g1=dd2[::-1]

 g2=g1[0:120]

 g3=g2[::-1]

 gg=DataFrame(g3)

 gg.T.to_excel('gg.xls') 

 #dd3=pd.DataFrame(dd2)
 #dd3.T.to_excel('d8.xls') 

 g=dd2[0:140]
 for i in range(dd.values.shape[0]-34):

     s=dd2[i*4:i*4+140]
     g=row_stack((g,s))

 fg=DataFrame(g)

 print(fg)
 fg.to_excel('fg.xls') 

 #-*- coding: utf-8 -*-
 #建立、训练多层神经网络，并完成模型的检验
 #from __future__ import print_function
 import pandas as pd

 inputfile1='fg.xls' #训练数据
 testoutputfile = 'test_output_data.xls' #测试数据模型输出文件
 data_train = pd.read_excel(inputfile1) #读入训练数据(由日志标记事件是否为洗浴)
 data_mean = data_train.mean()
 data_std = data_train.std()
 data_train1 = (data_train-data_mean)/5  #数据标准化

 y_train = data_train1.iloc[:,120:140].as_matrix() #训练样本标签列
 x_train = data_train1.iloc[:,0:120].as_matrix() #训练样本特征
 #y_test = data_test.iloc[:,4].as_matrix() #测试样本标签列

 from keras.models import Sequential
 from keras.layers.core import Dense, Dropout, Activation

 model = Sequential() #建立模型
 model.add(Dense(input_dim = 120, output_dim = 240)) #添加输入层、隐藏层的连接
 model.add(Activation('relu')) #以Relu函数为激活函数
 model.add(Dense(input_dim = 240, output_dim = 120)) #添加隐藏层、隐藏层的连接
 model.add(Activation('relu')) #以Relu函数为激活函数
 model.add(Dense(input_dim = 120, output_dim = 120)) #添加隐藏层、隐藏层的连接
 model.add(Activation('relu')) #以Relu函数为激活函数
 model.add(Dense(input_dim = 120, output_dim = 20)) #添加隐藏层、输出层的连接
 model.add(Activation('sigmoid')) #以sigmoid函数为激活函数
 #编译模型，损失函数为binary_crossentropy，用adam法求解
 model.compile(loss='mean_squared_error', optimizer='adam')

 model.fit(x_train, y_train, nb_epoch = 100, batch_size = 8) #训练模型
 model.save_weights('net.model') #保存模型参数

 inputfile2='gg.xls' #预测数据
 pre = pd.read_excel(inputfile2)                  

 pre_mean = data_mean[0:120]
 pre_std = pre.std()
 pre1 = (pre-pre_mean)/5  #数据标准化

 pre2 = pre1.iloc[:,0:120].as_matrix() #预测样本特征
 r = pd.DataFrame(model.predict(pre2))
 rt=r*5+data_mean[120:140].as_matrix()
 print(rt.round(2))

 rt.to_excel('rt.xls') 

 #print(r.values@data_train.iloc[:,116:120].std().values+data_mean[116:120].as_matrix())

 a=list(df.index[0:-1])

 b=a[0]

 c= datetime.datetime.strptime(b,'%Y-%m-%d')

 d = date2num(c)

 c1=[d+i+1 for i in range(5)]
 c2=np.array([c1])

 r1=rt.values.flatten()
 r2=r1[0:4]
 for i in range(4):

     r3=r1[i*4+4:i*4+8]
     r2=row_stack((r2,r3))

 c3=column_stack((c2.T,r2))
 r5=DataFrame(c3)

 if len(c3) == 0:
     raise SystemExit

 fig, ax = plt.subplots()
 fig.subplots_adjust(bottom=0.2)

 #ax.xaxis.set_major_locator(mondays)
 #ax.xaxis.set_minor_locator(alldays)
 #ax.xaxis.set_major_formatter(mondayFormatter)
 #ax.xaxis.set_minor_formatter(dayFormatter)

 #plot_day_summary(ax, quotes, ticksize=3)
 candlestick_ohlc(ax, c3, width=0.6, colorup='r', colordown='g')

 ax.xaxis_date()
 ax.autoscale_view()
 plt.setp(plt.gca().get_xticklabels(), rotation=45, horizontalalignment='right')

 ax.grid(True)
 #plt.title('000002')
 plt.show()