协方差/相关矩阵/相关系数

A = [a1, a2, ..., an]
B = [b1, b2, ..., bn]

ave_a = (a1 + a2 +...+ an)/n
ave_b = (b1 + b2 +...+ bn)/m

dev_a = [a1, a2, ..., an] - ave_a
dev_b = [b1, b2, ..., bn] - ave_b

cov_ab = ave(dev_a x dev_b)
cov_ba = ave(dev_b x dev_a)

import numpy as np
import matplotlib.pyplot as mp

a = np.random.randint(1, 30, 10)
b = np.random.randint(1, 30, 10)
#平均值
ave_a = np.mean(a)
ave_b = np.mean(b)
#离差
dev_a = a - ave_a
dev_b = b - ave_b
#协方差
cov_ab = np.mean(dev_a*dev_b)
cov_ba = np.mean(dev_b*dev_a)
print(‘a与b数组：‘, a, b)
print(‘a与b样本方差：‘, np.sum(dev_a**2)/(len(dev_a)-1), np.sum(dev_b**2)/(len(dev_b)-1))
print(‘a与b协方差：‘,cov_ab, cov_ba)
#绘图，查看两条图线的相关性
mp.figure(‘COV LINES‘, facecolor=‘lightgray‘)
mp.title(‘COV LINES‘, fontsize=16)
mp.xlabel(‘x‘, fontsize=14)
mp.ylabel(‘y‘, fontsize=14)
x = np.arange(0, 10)
#a,b两条线
mp.plot(x, a, color=‘dodgerblue‘, label=‘Line1‘)
mp.plot(x, b, color=‘limegreen‘, label=‘Line2‘)
#a,b两条线的平均线
mp.plot([0, 9], [ave_a, ave_a], color=‘dodgerblue‘, linestyle=‘--‘, alpha=0.7, linewidth=3)
mp.plot([0, 9], [ave_b, ave_b], color=‘limegreen‘, linestyle=‘--‘, alpha=0.7, linewidth=3)

mp.grid(linestyle=‘--‘, alpha=0.5)
mp.legend()
mp.tight_layout()
mp.show()

# a组样本 与 b组样本做对照后的相关系数
cov_ab/(std_a x std_b)
# b组样本 与 a组样本做对照后的相关系数
cov_ba/(std_b x std_a)
# a样本与a样本作对照   b样本与b样本做对照   二者必然相等
cov_ab/(std_a x std_b)=cov_ba/(std_b x std_a)

若相关系数越接近于0，越表示两组样本越不相关。
若相关系数越接近于1，越表示两组样本正相关。
若相关系数越接近于-1，越表示两组样本负相关。

print(‘相关系数：‘, cov_ab/(np.std(a)*np.std(b)), cov_ba/(np.std(a)*np.std(b)))

# 相关矩阵
numpy.corrcoef(a, b)    
# 相关矩阵的分子矩阵 
# [[a方差，ab协方差], [ba协方差, b方差]]
numpy.cov(a, b)        

# 协方差
import numpy as np
import matplotlib.pyplot as mp
import datetime as dt
import matplotlib.dates as md


def dmy2ymd(dmy):
  """
  把日月年转年月日
  :param day:
  :return:
  """
  dmy = str(dmy, encoding=‘utf-8‘)
  t = dt.datetime.strptime(dmy, ‘%d-%m-%Y‘)
  s = t.date().strftime(‘%Y-%m-%d‘)
  return s


dates, bhp_closing_prices =   np.loadtxt(‘bhp.csv‘,
             delimiter=‘,‘,
             usecols=(1, 6),
             unpack=True,
             dtype=‘M8[D],f8‘,
             converters={1: dmy2ymd})  # 日月年转年月日
vale_closing_prices =   np.loadtxt(‘vale.csv‘,
             delimiter=‘,‘,
             usecols=(6,),
             unpack=True)  # 因为日期一样,所以此处不读日期
# print(dates)
# 绘制收盘价的折现图
mp.figure(‘COV‘, facecolor=‘lightgray‘)
mp.title(‘COV‘, fontsize=18)
mp.xlabel(‘Date‘, fontsize=14)
mp.ylabel(‘Price‘, fontsize=14)
mp.grid(linestyle=":")

# 设置刻度定位器
# 每周一一个主刻度,一天一个次刻度

ax = mp.gca()
ma_loc = md.WeekdayLocator(byweekday=md.MO)
ax.xaxis.set_major_locator(ma_loc)
ax.xaxis.set_major_formatter(md.DateFormatter(‘%Y-%m-%d‘))
ax.xaxis.set_minor_locator(md.DayLocator())
# 修改dates的dtype为md.datetime.datetiem
dates = dates.astype(md.datetime.datetime)
mp.plot(dates, bhp_closing_prices,
        color=‘dodgerblue‘,
        linewidth=2,
        linestyle=‘-‘,
        alpha=0.8,
        label=‘BHP Closing Price‘)
mp.plot(dates, vale_closing_prices,
        color=‘orangered‘,
        linewidth=2,
        linestyle=‘-‘,
        alpha=0.8,
        label=‘VALE Closing Price‘)
# 计算两只股票收盘价的协方差
# 求均值
mean_bhp = np.mean(bhp_closing_prices)
mean_vale = np.mean(vale_closing_prices)
# 求离差
dev_bhp = bhp_closing_prices - mean_bhp
dev_vale = vale_closing_prices - mean_vale
# 协方差(bhp的离差乘以vale的离差)
cov = (dev_bhp * dev_vale).mean()  # 用的是总体的数据
print(‘cov:‘, cov)  # cov: 3.135577333333333
# 两条图线的相关性
k = cov / np.std(bhp_closing_prices * np.std(vale_closing_prices))
print(k)  # 0.8664988296368301
# 相关系数 = a方差/(a标准差*a标准差)
k = np.corrcoef(bhp_closing_prices, vale_closing_prices)
print(k)
"""
[[1.         0.86649883]
 [0.86649883 1.        ]]
"""
# q秋两组数据的协方差
covm = np.cov(bhp_closing_prices, vale_closing_prices)  # 用的是 样本的数据
print(covm)
"""
[[8.53844379 3.24370069]
 [3.24370069 1.64122023]]
"""

mp.legend()
mp.gcf().autofmt_xdate()
mp.show()