Numpy学习笔记（二）

# 导入numpy库

import numpy as np

?

# 文件读写

# 先创建一个单位矩阵

a = np.eye(2) # numpy自带的eye方法

print a

# 使用savetxt函数保存

np.savetxt("D:\Learn\Code\python\exercise\eye.txt", a)

# delimiter定义分隔符，默认是空格；usecols定义选取的；unpack默认False，解包

c, v = np.loadtxt(‘D:\Learn\Code\python\exercise\data.csv‘,

delimiter=",", usecols=(6, 7), unpack=True)

print u"收盘价:", c

print u"成交量:", v

c, v = np.loadtxt("D:\Learn\Code\python\exercise\data.csv",

delimiter=",", usecols=(6, 7), unpack=True)

# 计算以成交量加权的价格

vwap = np.average(c, weights=v)

print "VWAP:", vwap

?

# 算术平均价格

mean = np.mean(c)

print "Mean:", mean

?

# 时间加权平均价格

# 先构造一个时间序列

t = np.arange(len(c))

twap = np.average(c, weights=t)

print "TWAP:", twap

h, l = np.loadtxt("D:\Learn\Code\python\exercise\data.csv",

delimiter=",", usecols=(4, 5), unpack=True)

# 直接调用max()和min()函数

print "highest:", np.max(h)

print "lowest:", np.min(l)

?

# ptp函数可以计算取值范围

print "Spred high price:", np.ptp(h)

print "Spred low price:", np.ptp(l)

c = np.loadtxt("D:\Learn\Code\python\exercise\data.csv",

delimiter=",", usecols=(6, ), unpack=True)

# 寻找中位数

print "median:", np.median(c)

# 利用排序来检验正确与否

sorted_c = np.msort(c)

print "sorted_c:", sorted_c

N = len(c)

if N%2 != 0:

middle = sorted_c[N/2]

else:

middle = (sorted_c[N/2] + sorted_c[N/2-1])/2

print "middle:", middle

?

# 计算方差

variance = np.var(c)

print "variance:", variance

c = np.loadtxt("D:\Learn\Code\python\exercise\data.csv",

delimiter=",", usecols=(6, ), unpack=True)

# 简单收益率

# diff函数返回相邻数组元素差值组成的数组

diff = np.diff(c)

print "diff:", diff

# 收益率等于这一天与前一天的差值除以前一天的值

returns = diff/c[:-1]

print "returns:", returns

# 计算收益率的标注差

print "Standard deviation:", np.std(returns)

?

# 对数收益率

logreturns = np.diff(np.log(c))

print "logreturns:", logreturns

print "Standard deviation:", np.std(logreturns)

?

# 计算收益率为正值的情况

posretindices = np.where(returns > 0)

print "Indices with positive returns:", posretindices

?

# 波动率（对数收益率的标准差除以其均值，再除以交易日倒数的平方根，交易日通常取252天）

annual_volatility = np.std(logreturns)/np.mean(logreturns)/np.sqrt(1.0/252)

print "Annual volatility:", annual_volatility

print "Month volatility:", annual_volatility * np.sqrt(1.0/12)

# 先导入时间模块

import datetime

# 编写转换函数

def datetostr(s):

truetime = datetime.datetime.strptime(s, ‘%d-%m-%Y‘).weekday()

return truetime

# 开始读取数据，并添加converters参数

dates, close = np.loadtxt("D:\Learn\Code\python\exercise\data.csv",

delimiter=",", usecols=(1, 6), converters={1: datetostr},

unpack=True)

print "Dates:", dates

# 创建一个包含五个元素的数组

averages = np.zeros(5)

# where会根据条件返回满足条件的元素索引，take可以从索引中取出数据

for i in range(5):

indices = np.where(dates == i)

prices = np.take(close, indices)

avg = np.mean(prices)

print "Day", i, "prices", prices, "Average", avg

averages[i] = avg

import datetime

def datetostr(s):

truetime = datetime.datetime.strptime(s, ‘%d-%m-%Y‘).weekday()

return truetime

dates, open, high, low, close = np.loadtxt("D:\Learn\Code\python\exercise\data.csv",

delimiter=",", usecols=(1, 3, 4, 5, 6), converters={1: datetostr},

unpack=True)

# 为了方便计算，仅取前三周数据

dates = dates[:16]

?

# 寻找第一个星期一

first_monday = np.ravel(np.where(dates == 0))[0] # where返回的是个多维数组，需要展平

print "First Monday", first_monday

# 寻找最后一个星期五

last_friday = np.ravel(np.where(dates == 4))[-1]

print "Last Friday"

?

weeks_indices = np.arange(first_monday, last_friday+1)

print "Weeks indices initial", weeks_indices

?

weeks_indices = np.split(weeks_indices, 3)

print "Weeks indices after split", weeks_indices

?

# 为了后面的apply_along_axis

def summarize(a, o, h, l, c):

monday_open = o[a[0]]

week_high = np.max(np.take(h, a))

week_low = np.min(np.take(l, a))

friday_close = c[a[-1]]

return ("APPL", monday_open, week_high, week_low, friday_close)

# apply_along_axis内涵很丰富

weeksummary = np.apply_along_axis(summarize, 1, weeks_indices, open, high, low, close)

print "Week summary", weeksummary

# savetxt参数其实有很多

np.savetxt("D:\Learn\Code\python\exercise\weeksummary.csv", weeksummary, delimiter=",", fmt="%s")

h, l, c = np.loadtxt(‘D:\Learn\Code\python\exercise\data.csv‘, delimiter=‘,‘, usecols=(4, 5, 6), unpack=True)

?

N = 20

# 切片

h = h[-N:]

l = l[-N:]

?

print "len(h)", len(h), "len(l)", len(l)

print "Close", c

# 计算前一日的收盘价

previousclose = c[-N-1:-1]

?

print "len(previousclose)", len(previousclose)

print "Previous close", previousclose

# maximum函数可以选择出每个元素的最大值

truerange = np.maximum(h-l, h-previousclose, previousclose-l)

?

print "True range", truerange

# zeros函数初始化数组为0

atr = np.zeros(N)

?

atr[0] = np.mean(truerange)

?

for i in range(1, N):

atr[i] = (N-1)*atr[i-1] + truerange[i]

atr[i] /= N

?

print "ATR", atr

# N是移动窗口的大小

N = 5

# 权重是个平均值

weights = np.ones(N) / N

print "Weights", weights

?

c = np.loadtxt(‘D:\Learn\Code\python\exercise\data.csv‘, delimiter=‘,‘, usecols=(6, ), unpack=True)

# 要从卷积运算中取出与原数组重叠的区域

sma = np.convolve(weights, c)[N-1:-N+1]

# 生成一个时间序列

t = np.arange(N-1, len(c))

# 用matplotlib绘图

plt.plot(t, c[N-1:],‘b-‘, lw=1.0)

plt.plot(t, sma, ‘r-‘, lw=2.0)

plt.show()

plt.ylabel("Close Price")

plt.title(u"Simple Moving Average")

plt.annotate(‘before convolve‘, xy=(12.8, 363), xytext=(15, 363),

arrowprops=dict(facecolor=‘black‘,shrink=0.005))

plt.annotate(‘after convolve‘, xy=(15, 358), xytext=(17, 358),

arrowprops=dict(facecolor=‘black‘,shrink=0.005))

x = np.arange(5)

# 对x求指数，exp函数

print "Exp", np.exp(x)

# linspace函数实现等距分隔

print "Linspace", np.linspace(-1, 0 ,5)

N = 5

# 上面是两个示范，下面才是真的

weights = np.exp(np.linspace(-1, 0, N))

weights /= weights.sum()

print "Weights", weights

?

c = np.loadtxt(‘D:\Learn\Code\python\exercise\data.csv‘, delimiter=‘,‘, usecols=(6, ), unpack=True)

ema = np.convolve(weights, c)[N-1:-N+1]

t = np.arange(N-1, len(c))

plt.plot(t, c[N-1:],‘b-‘, lw=1.0)

plt.plot(t, ema, ‘r-‘, lw=2.0)

plt.show()