0. Backgound

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。

说白了，就是给一堆可投资券，求各种配比的组合下，相同return方差最小(或者相同方差return最大)的点集合。

更形象来说，就是在return-σ象限把投资组合的各种配比可能性都点在图上，左上方的edge就是这个有效前沿。

1. 导入模块

import pandas as pd

import numpy as np

import matplotlib.pyplot as plt

#用于正常显示中文标签

plt.rcParams["font.sans-serif"]=["SimHei"]

plt.rcParams["axes.unicode_minus"] = False

# 启动windpy接口

from WindPy import w

w.start()

2. 准备函数

def risk_free_rate(date):

r = w.wsd("SHIBOR6M.IR", "close,settle", date, date, "")

if r.ErrorCode !=0:

print("WIND数据读取有误，ErrorCode=" + str(r.ErrorCode))

sys.exit()

r = r.Data[0][0]/100

return r

def stock_close_data(stock_list,start_date,end_date):

stock_input = ",".join(stock_list)

data = w.wsd(stock_input, "close",start_date , end_date, "PriceAdj=F") #使用前复权处理

if data.ErrorCode!=0:

print("WIND数据读取有误，ErrorCode="+r.ErrorCode)

sys.exit()

print(data)

data = pd.DataFrame(data.Data,index = stock_list,columns = data.Times).T

return data

def generate_random_weight(num, stock_list):

weight = np.random.uniform(1,100,size = [num,len(stock_list)])

for i in range(0, num):

a_sum = 0.0

for j in range(0,len(stock_list)):

a_sum = a_sum +weight[i][j]

for k in range(0,len(stock_list)):

weight[i][k]=round(float(weight[i][k]/a_sum),4)

print("随机权重Ready！")

return weight

def portfolio_return(weight, mean, num, stock_list):

portfolio_re = np.random.uniform(0,0,size =num)

for i in range(0,num):

for j in range(0,len(stock_list)):

portfolio_re[i] = portfolio_re[i] + weight[i][j] * mean[j]

print("收益计算OKK！")

return portfolio_re

def portfolio_standard_deviation(weight, cov_matrix, num, stock_list):

portfolio_standard_dev = np.random.uniform(0,0,size=num)

for i in range(0, num):

var = np.dot(weight[i],cov_matrix)

var = np.dot(var,weight[i].T) # 组合方差

portfolio_standard_dev[i] = np.sqrt(var)

print("标准差搞定！")

return portfolio_standard_dev

def max_portfolio_return(p_return):

'''收益最大组合'''

p_return_list = list(p_return)

p_return_max = max(p_return_list)

p_return_max_index = p_return_list.index(p_return_max)

return p_return_max_index, p_return_max

def min_portfolio_std(p_std):

'''方差最小组合'''

p_std_list = list(p_std)

p_std_min = min(p_std_list)

p_std_min_index = p_std_list.index(p_std_min)

return p_std_min_index, p_std_min

def compute_sharp_ratio(returns, s_deviation, rfr):

'''

计算各组合的sharp_r = sharp_ratio_list, max_spr, max_spr_index

'''

sharp_ratio = np.random.uniform(0,0,len(returns))

for i in range(0,len(sharp_ratio)):

sharp_ratio[i] = (returns[i] - rfr) / s_deviation[i]

sharp_ratio_list = list(sharp_ratio)

return sharp_ratio, max(sharp_ratio_list), sharp_ratio_list.index(max(sharp_ratio_list))

def show_frontier(returns, s_deviation, max_return, min_std,sharp_r):

# 设置坐标轴的lable

plt.xlabel(" σ ")

plt.ylabel(" return % ")

colors1 = "#0080FF" #组合点的颜色

colors2 = "#CC0000" #最大收益点

colors3 = "#669900" #最小方差点

colors4 = "#FE9A2E" #最大sharpRatio

s = np.pi * 0.4 ** 2 # 普通组合点的大小

s1 = np.pi * 2 ** 2 # 最大瘦一点、最小方差点大小

plt.title("马科维茨有效前沿")

plt.scatter(s_deviation, returns, c=colors1, alpha=0.2,s=s)

plt.scatter(s_deviation[max_return[0]],max_return[1],c=colors2,alpha=0.8, s=s1, label = "最大收益")

plt.scatter(min_std[1],returns[min_std[0]],c=colors3,alpha=0.8,s=s1, label = "最小方差")

plt.scatter(s_deviation[sharp_r[2]],returns[sharp_r[2]],c=colors4,alpha = 0.8,s=s1, label = "最大SharpRatio")

plt.legend(loc = "best")

plt.plot()

plt.show()

return

3. 取数、作图

stock_list = ["600159.SH","000858.SZ","000615.SZ","300015.SZ","601318.SH"]

start_date = "2019-01-01"

end_date = "2020-04-23"

num = 50000 #生成多少个组合

# 计算一下无风险收益率

r = risk_free_rate("2020-04-23")

data = stock_close_data(stock_list,start_date,end_date)

daily_return = np.log(data/data.shift(1))

# daily_return = daily_return.shift(1)

mean = daily_return.mean() * 252 #年化

cov_matrix = daily_return.cov() * 252

weight = generate_random_weight(num, stock_list)

ptf_return = portfolio_return(weight,mean,num,stock_list)

ptf_std = portfolio_standard_deviation(weight, cov_matrix, num, stock_list)

# 最大收益率和最小方差

max_return= max_portfolio_return(ptf_return)

min_std=min_portfolio_std(ptf_std)

# 找Sharp Ratio

# sharp_r = sharp_ratio_list, max_spr, max_spr_index

sharp_r = compute_sharp_ratio(ptf_return, ptf_std, r)

show_frontier(ptf_return,ptf_std,max_return,min_std,sharp_r)

image.png

4. 总结

stock list乱拉了好多试，发现组合可选的券多起来以后，有效前沿的样子会没有那么完美，但是list短短的时候，样子可能会及其好看(然而并没有什么用处)

num=50000不变的情况下

比如只用3个股票的时候

image.png

用20个股票的时候

image.png

啊这个复杂又魔幻的世界= =