单因子——数据Wind
作者：互联网
读取数据
import time
date = time.strftime("%Y-%m-%d")
date = date.replace("'", "")
from WindPy import *
import pandas as pd
import numpy as np
from pymongo import MongoClient
w.start()
w.isconnected()
import time
from concurrent.futures import ProcessPoolExecutor,ThreadPoolExecutor, as_completed
import json
连接数据库
host ='xxx.xxx.x.xx'
port='27017'
user='xxxx'
code='utf8utf8'
uri= "mongodb://"+user+":"+code+"@"+host+":"+port+"/"+"?authMechanism=SCRAM-SHA-1"
client=MongoClient(uri)
db = client['CN_Stock']
collection = db['Wind_Close_Price']
获取Wind中A股公司列表
A_list = w.wset("sectorconstituent","date=2019-07-30;sectorid=a001010100000000",usedf=True)
A_list = A_list[1]#取dataframe部分
多线程
df = pd.DataFrame(columns = ['wind_code','close','volume','amt','total_shares'] )
df['wind_code'] = A_list['wind_code']
df_list = A_list.copy()
#定义多线程函数
def applyParallel(dfGrouped, func):
ppe = ThreadPoolExecutor(10)#设置线程池中最多能同时运行的线程数目
futures = []
results = []
for name, group in dfGrouped:
p = ppe.submit(func, group)#提交函数和变量，立即返回
futures.append(p)
for future in as_completed(futures):
r = future.result()
results.append(r)
return pd.concat(results)#合并表
#定义获取股票价格函数
def get_price(df_list):
stk = df_list['wind_code'].iloc[0]
collection = db['Wind_Data_Raw_Price_Daily']
Adjust_Price_Daily = pd.DataFrame()
try:
data2 = w.wsd(stk, "close,industry_csrc12_n", "2018-01-10", "2019-08-08","industryType=1;PriceAdj=F",usedf=True)
data2 = data2[1]#前复权价格
data2['wind_code']= stk
Adjust_Price_Daily = pd.concat([Adjust_Price_Daily,data2])
time.sleep(0.5)
print("done2_"+stk)
except:
print('error_no_record'+stk)
df['state']= 'error_no_record'
return Adjust_Price_Daily
df_list['rank'] = df_list.index
t1=time.time()
Adjust_Price_Daily_temp = applyParallel(df_list.groupby('rank'), get_price)
t2=time.time()
print(t2-t1)
Adjust_Price_Daily.index = range(len(Adjust_Price_Daily))
#某只股票的价格走势图
Adjust_Price_Daily[(Adjust_Price_Daily['wind_code'] == '000001.SZ')].plot(y=['CLOSE'])
collection.drop()
collection.insert_many(json.loads(Adjust_Price_Daily.T.to_json()).values())
清洗数据，标准化
Adjust_Price_Daily['date'] = Adjust_Price_Daily.index
Adjust_Price_Daily.dropna(axis=0, how='any', inplace=True)#去掉有空值的行
Adjust_Price = Adjust_Price_Daily.pivot( index='date',values='CLOSE', columns='wind_code')#将SQL格式转化为DataFrame格式
Adjust_Price.dropna(axis=1, how='any', inplace=True)
df_r = Adjust_Price.iloc[:,:]/Adjust_Price.iloc[:,:].shift(1)-1#获得每只股票每天的回报率
df_m = Adjust_Price.iloc/Adjust_Price.iloc.shift(30)#获得股票30天动量
df_m.dropna(axis=0, how='any', inplace=True)
df_r.dropna(axis=0, how='any', inplace=True)
将动量作为因子
data_all = pd.DataFrame()
for i in range(len(df_m)):
print (i)
df_1= df_m.iloc[i:i+1,:]
df_1 = df_1.T
df_1.columns = ['m_fz']
q0,q1,q2,q3,q4,q5 = df_1['m_fz'].quantile([0,0.2,0.4,0.6,0.8,1.0])#获取分位点
bins = [q0-1,q1, q2, q3, q4, q5+1]
labels=['1', '2', '3', '4', '5']
df_1 = pd.cut(df_1.m_fz, bins,right=True,labels=labels) #将股票分为五组
df_2= df_r.iloc[i+15:i+16,:]
df_2 = df_2.T
df = pd.concat([df_1,df_2],axis=1)#获得一个df，有组别和return值
data = df.groupby('m_fz').mean()
#加入5-1，4-2
data_51 = data.iloc[4]-data.iloc[0]
data_42 = data.iloc[3]-data.iloc[1]
data = data.append(data_51,ignore_index=True)
data = data.append(data_42,ignore_index=True)
data_all = pd.concat([data_all,data],axis = 1)
data_all.index = ['1','2','3','4','5','5-1','4-2']
data_all = data_all.T
pic = (1+data_all).cumprod().plot()#画图
接下来以vol波动率为因子
df_vol = df_r.rolling(30).std()*(252**0.5)#年化利率
df_vol.dropna(axis=0, how='any', inplace=True)
data_all = pd.DataFrame()
for i in range(len(df_vol)):
print (i)
df_1= df_vol.iloc[i+15:i+16,:]
df_1 = df_1.T
df_1.columns = ['v_fz']
q0,q1,q2,q3,q4,q5 = df_1['v_fz'].quantile([0,0.2,0.4,0.6,0.8,1.0])
bins = [q0-1,q1, q2, q3, q4, q5+1]
labels=['1', '2', '3', '4', '5']
df_1 = pd.cut(df_1.v_fz, bins,right=True,labels=labels)
df_2= df_r.iloc[i:i+1,:]
df_2 = df_2.T
df = pd.concat([df_1,df_2],axis=1)
data = df.groupby('v_fz').mean()
data_51 = data.iloc[4]-data.iloc[0]
data_42 = data.iloc[3]-data.iloc[1]
data = data.append(data_51,ignore_index=True)
data = data.append(data_42,ignore_index=True)
data_all = pd.concat([data_all,data],axis = 1)
data_all = data_all.T
data_all.columns = ['1','2','3','4','5','5-1','4-2']
data_all.head()
(1+data_all).cumprod().plot()#画图
加入行业对冲
df_Industry = pd.read_excel('首先有一个行业列表的excel.xlsx',encoding='utf-8')#读取行业，当然这个也是wind下的
df_Industry.columns=['wind_code','Industry']
df_Industry.dropna(axis=0, how='any', inplace=True)
#result = df_Industry.values.tolist()
Adjust_Price_Daily['date'] = Adjust_Price_Daily.index
Adjust_Price_Daily = pd.merge(Adjust_Price_Daily,df_Industry,on='wind_code')
Adjust_Price_Daily.dropna(axis=0, how='any', inplace=True)
#获取行业list
Industry = df_Industry['Industry'].unique()
Industry = Industry.tolist()
#根据行业分组，选取每个行业表现最好的一组
data_all = pd.DataFrame()
data_al = pd.DataFrame()
for ind in Industry:
print(ind)
df = Adjust_Price_Daily[Adjust_Price_Daily['Industry']==ind]
df = df.pivot( index='date',values='CLOSE', columns='wind_code')
df.dropna(axis=1, how='any', inplace=True)
df_r = df.iloc[:,:]/df.iloc[:,:].shift(1)-1
df_m = df/df.shift(30)
df_m.dropna(axis=0, how='any', inplace=True)
df_r.dropna(axis=0, how='any', inplace=True)
for i in range(len(df_m)):
print (i)
df_1= df_m.iloc[i:i+1,:]
df_1 = df_1.T
df_1.columns = ['m_fz']
if len(df_1)<5:
data = df_r.max
else:
q0,q1,q2,q3,q4,q5 = df_1['m_fz'].quantile([0,0.2,0.4,0.6,0.8,1.0])
bins = [q0-1,q1, q2, q3, q4, q5+1]
labels=['1', '2', '3', '4', '5']
df_1 = pd.cut(df_1['m_fz'], bins,right=True,labels=labels)
df_2= df_r.iloc[i+15:i+16,:]
df_2 = df_2.T
df_3 = pd.concat([df_1,df_2],axis=1)
data = df_3.groupby('m_fz').mean()
data = data.iloc[4]
data_al = pd.concat([data_all,data],axis=1)
data_all = pd.concat([data_all,data_al],axis=1)
data_all.index = ['1','2','3','4','5','5-1','4-2']
data_all = data_all.T
(1+data_all).cumprod().plot()
标签：df,数据,Price,Daily,因子,Adjust,iloc,data,Wind 来源： https://blog.csdn.net/weixin_42659943/article/details/99681117