神经网络python源码分享

时间:2022-09-15 09:00:35

神经网络的逻辑应该都是熟知的了,在这里想说明一下交叉验证

交叉验证方法:

神经网络python源码分享

看图大概就能理解了,大致就是先将数据集分成K份,对这K份中每一份都取不一样的比例数据进行训练和测试。得出K个误差,将这K个误差平均得到最终误差

这第一个部分是BP神经网络的建立

参数选取参照论文:基于数据挖掘技术的股价指数分析与预测研究_胡林林

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
import math
import random
import tushare as ts
import pandas as pd
random.seed(0)
def getData(id,start,end):
  df = ts.get_hist_data(id,start,end)
  DATA=pd.DataFrame(columns=['rate1', 'rate2','rate3','pos1','pos2','pos3','amt1','amt2','amt3','MA20','MA5','r'])
  P1 = pd.DataFrame(columns=['high','low','close','open','volume'])
  DATA2=pd.DataFrame(columns=['R'])
  DATA['MA20']=df['ma20']
  DATA['MA5']=df['ma5']
  P=df['close']
  P1['high']=df['high']
  P1['low']=df['low']
  P1['close']=df['close']
  P1['open']=df['open']
  P1['volume']=df['volume']
 
  DATA['rate1']=(P1['close'].shift(1)-P1['open'].shift(1))/P1['open'].shift(1)
  DATA['rate2']=(P1['close'].shift(2)-P1['open'].shift(2))/P1['open'].shift(2)
  DATA['rate3']=(P1['close'].shift(3)-P1['open'].shift(3))/P1['open'].shift(3)
  DATA['pos1']=(P1['close'].shift(1)-P1['low'].shift(1))/(P1['high'].shift(1)-P1['low'].shift(1))
  DATA['pos2']=(P1['close'].shift(2)-P1['low'].shift(2))/(P1['high'].shift(2)-P1['low'].shift(2))
  DATA['pos3']=(P1['close'].shift(3)-P1['low'].shift(3))/(P1['high'].shift(3)-P1['low'].shift(3))
  DATA['amt1']=P1['volume'].shift(1)/((P1['volume'].shift(1)+P1['volume'].shift(2)+P1['volume'].shift(3))/3)
  DATA['amt2']=P1['volume'].shift(2)/((P1['volume'].shift(2)+P1['volume'].shift(3)+P1['volume'].shift(4))/3)
  DATA['amt3']=P1['volume'].shift(3)/((P1['volume'].shift(3)+P1['volume'].shift(4)+P1['volume'].shift(5))/3)
  templist=(P-P.shift(1))/P.shift(1)
  tempDATA = []
  for indextemp in templist:
    tempDATA.append(1/(1+math.exp(-indextemp*100)))
  DATA['r'] = tempDATA
  DATA=DATA.dropna(axis=0)
  DATA2['R']=DATA['r']
  del DATA['r']
  DATA=DATA.T
  DATA2=DATA2.T
  DATAlist=DATA.to_dict("list")
  result = []
  for key in DATAlist:
    result.append(DATAlist[key])
  DATAlist2=DATA2.to_dict("list")
  result2 = []
  for key in DATAlist2:
    result2.append(DATAlist2[key])
  return result
def getDataR(id,start,end):
  df = ts.get_hist_data(id,start,end)
  DATA=pd.DataFrame(columns=['rate1', 'rate2','rate3','pos1','pos2','pos3','amt1','amt2','amt3','MA20','MA5','r'])
  P1 = pd.DataFrame(columns=['high','low','close','open','volume'])
  DATA2=pd.DataFrame(columns=['R'])
  DATA['MA20']=df['ma20'].shift(1)
  DATA['MA5']=df['ma5'].shift(1)
  P=df['close']
  P1['high']=df['high']
  P1['low']=df['low']
  P1['close']=df['close']
  P1['open']=df['open']
  P1['volume']=df['volume']
 
  DATA['rate1']=(P1['close'].shift(1)-P1['open'].shift(1))/P1['open'].shift(1)
  DATA['rate2']=(P1['close'].shift(2)-P1['open'].shift(2))/P1['open'].shift(2)
  DATA['rate3']=(P1['close'].shift(3)-P1['open'].shift(3))/P1['open'].shift(3)
  DATA['pos1']=(P1['close'].shift(1)-P1['low'].shift(1))/(P1['high'].shift(1)-P1['low'].shift(1))
  DATA['pos2']=(P1['close'].shift(2)-P1['low'].shift(2))/(P1['high'].shift(2)-P1['low'].shift(2))
  DATA['pos3']=(P1['close'].shift(3)-P1['low'].shift(3))/(P1['high'].shift(3)-P1['low'].shift(3))
  DATA['amt1']=P1['volume'].shift(1)/((P1['volume'].shift(1)+P1['volume'].shift(2)+P1['volume'].shift(3))/3)
  DATA['amt2']=P1['volume'].shift(2)/((P1['volume'].shift(2)+P1['volume'].shift(3)+P1['volume'].shift(4))/3)
  DATA['amt3']=P1['volume'].shift(3)/((P1['volume'].shift(3)+P1['volume'].shift(4)+P1['volume'].shift(5))/3)
  templist=(P-P.shift(1))/P.shift(1)
  tempDATA = []
  for indextemp in templist:
    tempDATA.append(1/(1+math.exp(-indextemp*100)))
  DATA['r'] = tempDATA
  DATA=DATA.dropna(axis=0)
  DATA2['R']=DATA['r']
  del DATA['r']
  DATA=DATA.T
  DATA2=DATA2.T
  DATAlist=DATA.to_dict("list")
  result = []
  for key in DATAlist:
    result.append(DATAlist[key])
  DATAlist2=DATA2.to_dict("list")
  result2 = []
  for key in DATAlist2:
    result2.append(DATAlist2[key])
  return result2
def rand(a, b):
  return (b - a) * random.random() + a
def make_matrix(m, n, fill=0.0):
  mat = []
  for i in range(m):
    mat.append([fill] * n)
  return mat
def sigmoid(x):
  return 1.0 / (1.0 + math.exp(-x))
def sigmod_derivate(x):
  return x * (1 - x)
class BPNeuralNetwork:
  def __init__(self):
    self.input_n = 0
    self.hidden_n = 0
    self.output_n = 0
    self.input_cells = []
    self.hidden_cells = []
    self.output_cells = []
    self.input_weights = []
    self.output_weights = []
    self.input_correction = []
    self.output_correction = []
 
  def setup(self, ni, nh, no):
    self.input_n = ni + 1
    self.hidden_n = nh
    self.output_n = no
    # init cells
    self.input_cells = [1.0] * self.input_n
    self.hidden_cells = [1.0] * self.hidden_n
    self.output_cells = [1.0] * self.output_n
    # init weights
    self.input_weights = make_matrix(self.input_n, self.hidden_n)
    self.output_weights = make_matrix(self.hidden_n, self.output_n)
    # random activate
    for i in range(self.input_n):
      for h in range(self.hidden_n):
        self.input_weights[i][h] = rand(-0.2, 0.2)
    for h in range(self.hidden_n):
      for o in range(self.output_n):
        self.output_weights[h][o] = rand(-2.0, 2.0)
    # init correction matrix
    self.input_correction = make_matrix(self.input_n, self.hidden_n)
    self.output_correction = make_matrix(self.hidden_n, self.output_n)
 
  def predict(self, inputs):
    # activate input layer
    for i in range(self.input_n - 1):
      self.input_cells[i] = inputs[i]
    # activate hidden layer
    for j in range(self.hidden_n):
      total = 0.0
      for i in range(self.input_n):
        total += self.input_cells[i] * self.input_weights[i][j]
      self.hidden_cells[j] = sigmoid(total)
    # activate output layer
    for k in range(self.output_n):
      total = 0.0
      for j in range(self.hidden_n):
        total += self.hidden_cells[j] * self.output_weights[j][k]
      self.output_cells[k] = sigmoid(total)
    return self.output_cells[:]
  def back_propagate(self, case, label, learn, correct):
    # feed forward
    self.predict(case)
    # get output layer error
    output_deltas = [0.0] * self.output_n
    for o in range(self.output_n):
      error = label[o] - self.output_cells[o]
      output_deltas[o] = sigmod_derivate(self.output_cells[o]) * error
    # get hidden layer error
    hidden_deltas = [0.0] * self.hidden_n
    for h in range(self.hidden_n):
      error = 0.0
      for o in range(self.output_n):
        error += output_deltas[o] * self.output_weights[h][o]
      hidden_deltas[h] = sigmod_derivate(self.hidden_cells[h]) * error
    # update output weights
    for h in range(self.hidden_n):
      for o in range(self.output_n):
        change = output_deltas[o] * self.hidden_cells[h]
        self.output_weights[h][o] += learn * change + correct * self.output_correction[h][o]
        self.output_correction[h][o] = change
    # update input weights
    for i in range(self.input_n):
      for h in range(self.hidden_n):
        change = hidden_deltas[h] * self.input_cells[i]
        self.input_weights[i][h] += learn * change + correct * self.input_correction[i][h]
        self.input_correction[i][h] = change
    # get global error
    error = 0.0
    for o in range(len(label)):
      error += 0.5 * (label[o] - self.output_cells[o]) ** 2
    return error
  def train(self, cases, labels, limit=10000, learn=0.05, correct=0.1):
    for i in range(limit):
      error = 0.0
      for i in range(len(cases)):
        label = labels[i]
        case = cases[i]
        error += self.back_propagate(case, label, learn, correct)
  def test(self,id):
    result=getData("000001", "2015-01-05", "2015-01-09")
    result2=getDataR("000001", "2015-01-05", "2015-01-09")
    self.setup(11, 5, 1)
    self.train(result, result2, 10000, 0.05, 0.1)
    for t in resulttest:
      print(self.predict(t))

下面是选取14-15年数据进行训练,16年数据作为测试集,调仓周期为20个交易日,大约1个月,对上证50中的股票进行预测,选取预测的涨幅前10的股票买入,对每只股票分配一样的资金,初步运行没有问题,但就是太慢了,等哪天有空了再运行

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
import BPnet
import tushare as ts
import pandas as pd
import math
import xlrd
import datetime as dt
import time
 
#
#nn =BPnet.BPNeuralNetwork()
#nn.test('000001')
#for i in ts.get_sz50s()['code']:
holdList=pd.DataFrame(columns=['time','id','value'])
share=ts.get_sz50s()['code']
time2=ts.get_k_data('000001')['date']
newtime = time2[400:640]
newcount=0
for itime in newtime:
  print(itime)
  if newcount % 20 == 0:
        sharelist = pd.DataFrame(columns=['time','id','value'])
    for ishare in share:
      backwardtime = time.strftime('%Y-%m-%d',time.localtime(time.mktime(time.strptime(itime,'%Y-%m-%d'))-432000*4))
      trainData = BPnet.getData(ishare, '2014-05-22',itime)
      trainDataR = BPnet.getDataR(ishare, '2014-05-22',itime)
      testData = BPnet.getData(ishare, backwardtime,itime)
      try:
        print(testData)
        testData = testData[-1]
        print(testData)
        nn = BPnet.BPNeuralNetwork()
        nn.setup(11, 5, 1)
        nn.train(trainData, trainDataR, 10000, 0.05, 0.1)
        value = nn.predict(testData)
        newlist= pd.DataFrame({'time':itime,"id":ishare,"value":value},index=["0"])
        sharelist = sharelist.append(newlist,ignore_index=True)
      except:
        pass
    sharelist=sharelist.sort(columns ='value',ascending=False)
    sharelist = sharelist[:10]
    holdList=holdList.append(sharelist,ignore_index=True)
  newcount+=1
  print(holdList)

总结

以上就是本文关于神经网络python源码分享的全部内容,希望对大家有所帮助。如有不足之处,欢迎留言指出。感谢朋友们对本站的支持!

原文链接:https://www.cnblogs.com/yunerlalala/p/6240722.html