forked from wuwangchuxinybf/multiple_factors_models
-
Notifications
You must be signed in to change notification settings - Fork 0
/
rotation_model.py
125 lines (90 loc) · 5.2 KB
/
rotation_model.py
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
# -*- coding: utf-8 -*-
"""
Created on Mon Apr 1 21:49:45 2019
@author: wuwangchuxin
"""
import pandas as pd
import numpy as np
from single_factors_test import Clean_Data
add_winddata = 'C:/Users/wuwangchuxin/Desktop/TF_SummerIntern/MF_data/wind/'
add_ready = 'C:/Users/wuwangchuxin/Desktop/TF_SummerIntern/MF_data/prepared_data/del_cixin/'
class Logistic_Regression:
def __init__(self,X,y,alpha):
self.X = X
self.y = y
self.alpha = alpha
@staticmethod
def sigmoid(z):
return 1 / (1 + np.exp(-z))
def gradient_function(self,theta):
'''Gradient of the function J definition.'''
h = self.sigmoid(np.dot(self.X,theta))
return np.dot(self.X.transpose(),h-self.y[:,np.newaxis])
def grad_descent(self):
'''梯度下降算法'''
theta = np.ones((self.X.shape[1], 1)) #初始化回归系数（n, 1) np.array([1, 1]).reshape(2, 1)
gradient = self.gradient_function(theta)
n=1
# while not np.all(np.absolute(gradient) <= 1e-5):
maxCycle = 100000
for i in range(maxCycle):
theta = theta - self.alpha*gradient #梯度
gradient = self.gradient_function(theta)
n+=1
print (n)
return theta
class Logit_style:
'''使用逻辑回归实现风格轮动'''
def __init__(self):
self.market_hs300 = pd.read_csv(add_winddata+'market_data_hs300.csv')
self.market_zz1000 = pd.read_excel(add_winddata+'market_data_zz1000.xlsx').iloc[:115,:]
self.market_exchange = pd.read_excel(add_winddata+'market_exchange.xlsx').iloc[:115,:]
self.market_M2 = pd.read_excel(add_winddata+'market_M2.xls').iloc[:115,:]
self.market_CPI = pd.read_excel(add_winddata+'market_CPI.xls').iloc[:115,:]
self.market_hs300_pettm = pd.read_excel(add_winddata+'market_hs300_pettm.xlsx').iloc[:115,:]
self.market_zz500_pettm = pd.read_excel(add_winddata+'market_zz500_pettm.xlsx').iloc[:115,:]
self.market_hs300_pblf = pd.read_excel(add_winddata+'market_hs300_pblf.xlsx').iloc[:115,:]
self.market_zz500_pblf = pd.read_excel(add_winddata+'market_zz500_pblf.xlsx').iloc[:115,:]
self.market_zzgz = pd.read_excel(add_winddata+'market_zzgz.xlsx').iloc[:115,:]
self.market_zzqz = pd.read_excel(add_winddata+'market_zzqz.xlsx').iloc[:115,:]
self.market_hs300_vol = pd.read_excel(add_winddata+'market_hs300_vol.xlsx').iloc[:115,:]
self.market_zz500_vol = pd.read_excel(add_winddata+'market_hs300_vol.xlsx').iloc[:115,:]
def gene_y(self):
'''生成0-1序列作为被解释变量，如果沪深300指数涨跌幅大于中证1000指数，则为1，否则为0'''
classLabels = np.array(self.market_hs300['chg']>self.market_zz1000['chg']).astype(int)[3:]
return classLabels
def gene_factors(self):
'''滞后一期：汇率变化，PE 差，PB 差，股债收益溢价，指数波动率差
滞后二期：货币供应量，消费价格水平
滞后一期和三期：大小盘指数涨幅差滞后项'''
fac = np.array([self.market_exchange['chg'][2:-1], #汇率变化;人民币兑美元汇率月度变化率;滞后1期
self.market_M2['chg'][1:-2],#货币供应量;M2;月同比增速 滞后2期
self.market_CPI['chg'][1:-2], #消费价格水平;CPI;月同比增速;滞后2期
#PE差;小盘指数滚动市盈率-大盘指数滚动市盈率;滞后1期
self.market_zz500_pettm['pe_ttm'][2:-1]-self.market_hs300_pettm['pe_ttm'][2:-1],
#PB差;小盘指数市净率-大盘指数市净率;滞后1期
self.market_zz500_pblf['chg'][2:-1]-self.market_hs300_pblf['chg'][2:-1],
#股债收益溢价;中证全指月涨幅-中证国债指数月涨幅;滞后1期
self.market_zzqz['chg'][2:-1]-self.market_zzgz['chg'][2:-1],
#指数波动率差;小盘股指数历史波动率-大盘股指数历史波动率;滞后1期
self.market_zz500_vol['chg'][2:-1]-self.market_hs300_vol['chg'][2:-1],
#大小盘指数涨幅差滞后项;滞后1期的大小盘指数月涨幅差;滞后1期
self.market_hs300['chg'][2:-1]-self.market_zz1000['chg'][2:-1],
#大小盘指数涨幅差滞后项;滞后3期的大小盘指数月涨幅差;滞后3期
self.market_hs300['chg'][:-3]-self.market_zz1000['chg'][:-3]]).T
dataMatIn = np.insert(fac, 0, 1, axis=1)
return dataMatIn
if __name__=='__main__':
alpha = 0.01
def Data_preposs(f_arr):
'''完整的数据预处理流程'''
f_a1 = Clean_Data(f_arr).Median_deextremum()
f_a2 = Clean_Data(f_a1).Z_score()
f_a4 = np.around(f_a2,decimals=4)
return f_a4
y = Logit_style().gene_y()
X = Logit_style().gene_factors()
X = Data_preposs(X)
theta_fit = Logistic_Regression(X,y,alpha).grad_descent()
h_fit = np.around(Logistic_Regression.sigmoid(np.dot(X,theta_fit)),4)
np.save(add_ready+'style_rotation_res',h_fit)