基于tushare数据构造有效前沿曲线

本文基于tushare数据构造有效前沿曲线（个人的tushareID：498867）

有效前沿是在给定投资范围，return-risk约束条件下同等风险情况下收益最大的的资产配置集合。

# Provides ways to work with large multidimensional arrays
import numpy as np 
# Allows for further data manipulation and analysis
import pandas as pd
from pandas_datareader import data as web # Reads stock data 
import matplotlib.pyplot as plt # Plotting
import matplotlib.dates as mdates # Styling dates
%matplotlib inline

import datetime as dt # For defining dates
import mplfinance as mpf # Matplotlib finance

import time

# Used to get data from a directory
import os
from os import listdir
from os.path import isfile, join

#Statsmodels is a great library we can use to run regressions.
import statsmodels.api as sm
# Seaborn extends the capabilities of Matplotlib
import seaborn as sns
# Used for calculating regressions
from statsmodels.tsa.ar_model import AutoReg, ar_select_order

name=['600418.SH', '603355.SH', '002429.SZ', '002945.SZ', '002423.SZ',
       '300376.SZ', '000656.SZ', '600985.SH', '600392.SH', '002085.SZ',
       '601106.SH', '600188.SH', '002958.SZ', '601611.SH', '688321.SH',
       '300724.SZ', '603267.SH', '002152.SZ', '603056.SH', '600968.SH']

# 导入tushare
import tushare as ts
import pandas as pd
dff=pd.DataFrame()
# 初始化pro接口
pro = ts.pro_api('你的token')
for i in name:
    # 导入tushare
    # 拉取数据
    try:
        df = pro.daily(**{
            "ts_code": i,
            "trade_date": "",
            "start_date": "20200101",
            "end_date": "20220101",
            "offset": "",
            "limit": ""
        }, fields=[
            "ts_code",
            "trade_date",
            "open",
            "high",
            "low",
            "close",
            "pre_close",
            "change",
            "pct_chg",
            "vol",
            "amount"
        ])
        dff[i]=df['pct_chg'].values
    except:
        pass

dff['trade_date']=df['trade_date'].values
dff=dff.sort_values('trade_date')
dff=dff.set_index('trade_date')

port_list = list(dff.columns)
mean_ret = dff.mean() * 252 # 252 average trading days per year
dff.cov() * 252

weights = np.random.random(len(port_list))
weights /= np.sum(weights)
print('Weights :', weights)
print('Total Weight :', np.sum(weights))

np.sum(weights * mean_ret) 
np.sqrt(np.dot(weights.T, np.dot(dff.cov() * 252, weights)))

p_ret = [] # Returns list
p_vol = [] # Volatility list
p_SR = [] # Sharpe Ratio list
p_wt = [] # Stock weights list
risk_free_rate = 0.0125
n=len(port_list)

for x in range(40000):
    # Generate random weights
    p_weights = np.random.random(n)
    p_weights /= np.sum(p_weights)
    
    # Add return using those weights to list
    ret_1 = np.sum(p_weights * mean_ret) * 252
    p_ret.append(ret_1)
    
    # Add volatility or standard deviation to list
    vol_1 = np.sqrt(np.dot(p_weights.T, np.dot(dff.cov() * 252, p_weights)))
    p_vol.append(vol_1)
    
    # Get Sharpe ratio
    SR_1 = (ret_1 - risk_free_rate) / vol_1
    p_SR.append(SR_1)
    
    # Store the weights for each portfolio
    p_wt.append(p_weights)
    
# Convert to Numpy arrays
p_ret = np.array(p_ret)
p_vol = np.array(p_vol)
p_SR = np.array(p_SR)
p_wt = np.array(p_wt)

p_ret, p_vol, p_SR, p_wt

# Create a dataframe with returns and volatility
ports = pd.DataFrame({'Return': p_ret, 'Volatility': p_vol})

ports.plot(x='Volatility', y='Return', kind='scatter', figsize=(16, 9))

最后得到如下图：