In [1]:
#!/Tsan/bin/python
# -*- coding: utf-8 -*-
In [125]:
# Libraries to use
from __future__ import division
import numpy as np
import pandas as pd
import statsmodels.api as sm
import matplotlib.pyplot as plt
import seaborn as sns
from datetime import datetime
import mysql.connector
import json
import collections
In [127]:
import talib as tb
In [4]:
# Import My own library for factor testing
from SingleFactorTest import factorFilterFunctions as ff
#from config import *
In [5]:
%matplotlib inline
%load_ext line_profiler
In [ ]:
In [6]:
# {'600036':u'招商银行','002142':u'宁波银行','601939':u'建设银行','000709':u'河钢股份','000807':u'云铝股份','600111':u'北方稀土',
# '002001':u'新和成','002562':u'兄弟科技','0300121':u'阳谷华泰','600352':u'浙江龙盛','600409':u'三友化工','000636':u'风华高科',
# '000962':u'东方钽业','300228':u'富瑞特装'，'600681':u'百川能源','600789':u'鲁抗医药','002166':u'莱茵生物','002166':u'万华化学',
# '000933':u'神火股份','600977':u'中国电影','000802':u'北京文化','600809':u'山西汾酒','600438':u'通威股份','000970':u'中科三环'}
In [7]:
# 股票推荐字典
stkdict = {'20170815':['600036','002142','601939'], '20170814':['000709','000807','600111','002001','002562'],\
'20170810':['300121','600352','600409'], '20170809':['000636','000962'], '20170808':['600803','300228','600681'],
'20170807':['600789','002166','600309'],'20170803':['000933','600977','000802','600809'],
'20170801':['600438'], '20170731':['600438','600111','000970']}
In [8]:
sortedstkdict = collections.OrderedDict(sorted(stkdict.items()))
In [9]:
fileNameClose = 'LZ_CN_STKA_QUOTE_TCLOSE.h5.csv'
fileNameOpen = 'LZ_CN_STKA_QUOTE_TOPEN.h5.csv'
In [10]:
path = ff.data_path
# Constants
startTime = datetime.strptime('20170731', '%Y%m%d')
endTime = datetime.strptime('20170816', '%Y%m%d')
In [11]:
# 开盘价
openPrice= pd.read_csv(path+fileNameOpen,infer_datetime_format=True,parse_dates=[0],index_col=0).loc[startTime:endTime]
stkList = map(lambda x: x.split('.')[0],openPrice.columns[1:])
openPrice = openPrice.iloc[:,:-1]
openPrice.columns = stkList
In [ ]:
# 收盘价
closePrice= pd.read_csv(path+fileNameClose,infer_datetime_format=True,parse_dates=[0],index_col=0).loc[startTime:endTime]
stkList = map(lambda x: x.split('.')[0],closePrice.columns[1:])
closePrice = closePrice.iloc[:,:-1]
closePrice.columns = stkList
In [ ]:
closePrice
In [ ]:
InitiCap = 100000 # 初始资金
capital = InitiCap
feeRate = 0.0003 # 手续费+冲击成本
stampTaxRate = 0.001 # 印花税
buyList = []
availableCash = {} # 每天的现金流量
# 生成购买的股票池
for key,item in sortedstkdict.iteritems():
print key
date = datetime.strptime(key, '%Y%m%d')
totalCostToday = 0 # 当天买入的所有股票的成本
for stk in item:
buyPrice = round(openPrice.loc[date][stk],2)
amount = 100
cost = round(buyPrice*amount*(1+feeRate),2)
buyList.append([date,stk,buyPrice,amount,cost,1]) # 日期，股票代码，购买价格，购买数量，总cost,方向
totalCostToday+= cost
assert capital > totalCostToday
capital = capital - totalCostToday
availableCash[date] = capital # 买入股票后剩下的资金
In [ ]:
In [ ]:
availableCash
In [ ]:
dateList = sorted([datetime.strptime(key, '%Y%m%d') for key in stkdict.keys()])
dateList = closePrice.loc[dateList[0]:dateList[-1]].index
dateList
In [ ]:
availableCash
In [ ]:
availableCashSeries = pd.Series(index = dateList,data = availableCash )
availableCashSeries = availableCashSeries.ffill()
availableCashSeries
In [ ]:
availableCashSeries
In [ ]:
buyList
In [ ]:
# 简单回测
cutwinthreshold = 0.05
cutlossthreshold = -0.05
pnldict = {}
sellList = []
for tradeitem in buyList:
pnl = (closePrice.loc[tradeitem[0]:][tradeitem[1]] - tradeitem[2]) / tradeitem[2] # 计算每天的pnl
cutwin = pnl[pnl >= cutwinthreshold]
cutloss = pnl[pnl <= cutlossthreshold]
if len(cutwin) > 0: # 如果达到止盈线，则按当天收盘价止盈
pnldict[tradeitem[1]] = round(cutwin .iloc[0],2)
selldate = cutwin.index[0]
sellPrice = closePrice.loc[selldate][tradeitem[1]]
cost = round(sellPrice*tradeitem[3]*(1-stampTaxRate),2)
sellList.append([selldate,tradeitem[1],round(sellPrice,2),tradeitem[3],cost,-1]) # 卖出list的格式与买入相同
availableCashSeries.loc[selldate:]+= cost # 记录到资金账户里
elif len(cutloss) > 0: # 反之则止损
pnldict[tradeitem[1]] = round(cutloss .iloc[0],2)
selldate = cutloss.index[0]
sellPrice = closePrice.loc[selldate][tradeitem[1]]
cost = round(sellPrice*tradeitem[3]*(1-stampTaxRate),2)
sellList.append([selldate,tradeitem[1],round(sellPrice,2),tradeitem[3],cost,-1]) # 卖出list的格式与买入相同
availableCashSeries.loc[selldate:]+= cost
else: # 没有达到止盈止损线的情况下，按照倒数第二天的收盘价计算pnl
pnldict[tradeitem[1]] = round(pnl.iloc[-2],2)
In [ ]:
sellList
In [ ]:
holdingList = list(set([i[1] for i in buyList]) - set([i[1] for i in sellList]))
In [ ]:
holdingList
In [ ]:
In [ ]:
selldatepair = {i[1]:i[0] for i in buyList if i[1] in holdingList}
selldatepair
In [ ]:
tempo = closePrice[holdingList]
In [ ]:
position = tempo.copy()
for stk,date in selldatepair.iteritems():
filterCondition = tempo[tempo[stk].index >= date]
falseCondition= tempo[tempo[stk].index < date]
position.loc[filterCondition.index,stk]= 100
position.loc[falseCondition.index,stk]= 0
position
In [ ]:
marketValue = position*tempo
In [ ]:
marketValue.sum(axis=1)+availableCashSeries
In [ ]:
# 计算卖出时得到的资金
for sellitem in sellList:
availableCashSeries.loc[sellitem[0]]+= sellitem[-2]
availableCashSeries
In [ ]:
# 合并买卖订单
tradingList = buyList + sellList
tradingCode = list(set([i[1] for i in buyList]) | set([i[1] for i in sellList]))
In [ ]:
tradingList
In [ ]:
# 创建tradingsheet
tradingSheet = pd.DataFrame(index = dateList, columns = tradingCode, data=0, dtype = float)
for item in tradingList:
tradingSheet .loc[item[0],item[1]]+= item[-1] * item[-3]
positionchangeDF = tradingSheet .cumsum().ffill()
In [ ]:
tradingSheet
In [ ]:
positionchangeDF
In [ ]:
# 总资产
totalCapital= (positionchangeDF * closePrice[positionchangeDF.columns]).sum(axis=1)+ availableCashSeries
In [ ]:
totalCapital
In [ ]:
# 年化收益率
annualizedRet = ((totalCapital.iloc[-2] - InitiCap)/ InitiCap +1)**(250/(len(totalCapital)-1)) -1
In [ ]:
annualizedRet
In [ ]:
availableCashSeries
In [ ]:
pnlseries = pd.Series(pnldict)
In [ ]:
pnlseries.describe()
In [ ]:
plt.hist(pnlseries)
In [ ]:
#-----------------------------------------------------Market Timing------------------------------------
In [ ]:
filenameHS300Member = 'LZ_CN_STKA_INDEX_HS300MEMBER.h5'
filenameCSI500Member = 'LZ_CN_STKA_INDEX_CSI500MEMBER.h5'
# 复权后价格
filenameAdjClose = 'OwnfactorAdjustedClose.h5'
In [ ]:
teststartTime = datetime.strptime('20120731', '%Y%m%d')
testendTime = datetime.strptime('20171016', '%Y%m%d')
In [ ]:
def getMemberList(path,filename):
df = ff.readh5data(path,filename).iloc[-1]
return df.loc[df==1].index.tolist()
In [ ]:
hs300MemberList = getMemberList(path, filenameHS300Member)
csi500MemberList = getMemberList(path, filenameCSI500Member)
In [ ]:
priceData = ff.readh5data(path,filenameAdjClose).pct_change().dropna(how='all').loc[teststartTime:testendTime]
In [ ]:
priceData[hs300MemberList].loc[datetime.strptime('20170810', '%Y%m%d')].name
In [ ]:
priceData[hs300MemberList].loc[datetime.strptime('20170810', '%Y%m%d')].hist()
In [ ]:
priceData[csi500MemberList].iloc[-2].name
In [ ]:
#-----------------------------------------------------CTA Backtest Engine------------------------------------
In [ ]:
def getNewMatrix(inputArray, t, m):
newMatrix = []
n = t-m+1
for i in range(n):
newdata = list(inputArray[i:m+i])
newMatrix.append(newdata)
#newMatrix = np.array(newMatrix).reshape(n,m)
return np.array(newMatrix)
def recreateArray(newMatrix,t,m):
ret = []
n = t - m + 1
for p in range(1, t+1):
if p < m:
alpha = p
elif p > t-m+1:
alpha = t-p+1
else:
alpha = m
sigma = 0
for j in range(1, m+1):
i = p - j + 1
if i > 0 and i < n+1:
sigma += newMatrix[i-1][j-1]
ret.append(sigma/alpha)
return np.array(ret)
def getSVD(inputArray,t,m):
#print 1
inputmatrix = getNewMatrix(inputArray, t, m)
u, s, v = np.linalg.svd(inputmatrix)
eviNum = 1 if s[0]/s.sum() > 0.99 else 2
sNew = np.zeros((eviNum, eviNum))
np.fill_diagonal(sNew, s[:eviNum])
matrixForts = np.dot(np.dot(u[:, :eviNum].reshape(u.shape[0], eviNum), sNew), v[:eviNum])
#print matrixForts.shape
newts = recreateArray(matrixForts, t, m)
return newts
In [12]:
old_path = 'C:/Users/LZJF_02/Desktop/original_data/'
new_path = 'C:/Users/LZJF_02/Desktop/modified_data/'
filename = 'rb000_1min.csv'
#newname = 'j9000_1min_modi.csv'
In [13]:
startTime = datetime.strptime('20160110', '%Y%m%d') # 上涨趋势
endTime = datetime.strptime('20161107', '%Y%m%d')
startTrainTime = datetime.strptime('20150530', '%Y%m%d')
endTimeTrain = datetime.strptime('20160530', '%Y%m%d')
startTimeVal = datetime.strptime('20151016', '%Y%m%d') # 下跌趋势
endTimeVal = datetime.strptime('20151123', '%Y%m%d')
startTimeSurge = datetime.strptime('20170606', '%Y%m%d') # 上涨趋势
endTimeSurge = datetime.strptime('20170809', '%Y%m%d')
startTimePlunge = datetime.strptime('20170906', '%Y%m%d') # 下跌趋势
endTimePlunge = datetime.strptime('20170929', '%Y%m%d')
In [14]:
# svd滤波参数
shapeNum = 30
svdShort = 7
svdLong = 20
In [15]:
# 期货合约相关参数
Leverage = 10
Slipage = 2
ContractSize = 10
IniCapital = 100000
FeeRate = 0.003
PriceTick = 1
In [16]:
data = pd.read_csv(old_path+filename,infer_datetime_format=True,header=None,names = ['Open', 'High', 'Low', 'Close','TotalVolume','OpenInterest'])
data.index = data.index.map(lambda x : pd.to_datetime(x))
In [17]:
data['OpenRatio'] = (data['OpenInterest'] - data['OpenInterest'].shift(1)) / data['TotalVolume']
In [18]:
data.loc[startTime:].describe()
data.loc[startTime:].quantile(0.85)
Out[18]:
In [19]:
data.tail(20)
Out[19]:
In [20]:
data.describe()
Out[20]:
In [21]:
d = {'x': {'b': 10, 'c': 20}, 'y': {'b': '33', 'c': 44,'a':1}}
In [22]:
data['TotalVolume'].idxmax()
Out[22]:
In [23]:
dd=[1,2,3,4.5,-2.3]
s =pd.Series(dd)
s[s<0].shape
Out[23]:
In [24]:
#pd.DataFrame.from_dict(f,orient='index')
In [176]:
c = u'1,231'.encode('utf-8').split(',')
float(c[0] +c[1])
Out[176]:
In [179]:
float("123,456.908".replace(',',''))
Out[179]:
In [25]:
dataIndice = data.loc[startTime:endTime]
indexby5MIN = filter(lambda x : x.minute % 5 ==0, dataIndice.index)
# modify data by customized method
#dataIndice = data.loc[startTime:]
resultList = []
for index, time in enumerate(indexby5MIN):
if index < len(indexby5MIN) - 1:
tempdata = dataIndice.loc[time:indexby5MIN [index+1]].iloc[:-1]
resultdict = {}
resultdict['Open'] = tempdata['Open'].values[0]
resultdict['High'] = tempdata['High'].max()
resultdict['Low'] = tempdata['Low'].min()
resultdict['Close'] = tempdata['Close'].values[-1]
resultdict['TotalVolume'] = tempdata['TotalVolume'].sum()
resultdict['OpenInterest'] = tempdata['OpenInterest'].values[-1]
resultdict['OpenRatio'] = tempdata['OpenRatio'].values[-1]
resultdict['OpenRatioModi'] = (np.sqrt(tempdata['TotalVolume']) * tempdata['OpenRatio'] / \
(np.sqrt(tempdata['TotalVolume']).sum())).mean()
resultdict['time'] = time
resultList.append(resultdict)
df = pd.DataFrame(resultList,columns = ['Open','High','Low','Close','TotalVolume','OpenInterest','OpenRatio','time','OpenRatioModi'])
dataClean = df.set_index('time')
dataClean['pnl'] = dataClean['Close'].pct_change()
dataClean['nextpnl'] = dataClean['pnl'].shift(1)
In [133]:
np_real_data = np.array(dataClean['Close'].values,dtype='f8')
In [140]:
alpha1 = (np.cos(0.707 * np.pi /48) + np.sin(0.707 * np.pi /48) - 1 ) / np.cos(0.707 * np.pi /48)
alpha1
Out[140]:
In [146]:
LPPeriod = 20
a1 = np.exp(-1.414 * np.pi / LPPeriod)
b1 = 2 * a1 * np.cos(1.414 * np.pi / LPPeriod)
c2 = b1
c3 = - a1 * a1
c1 = 1 - c2 - c3
In [ ]:
HP = np.zeros(close)
HP = (1 - alpha1 / 2) ** 2 * (Close - 2 * Close[-1] + Close[-2]) + 2 * (1 - alpha1) * HP[-1] - (1 - alpha1) ** 2 * HP[-2]
In [154]:
# Stay long if price is trading above HT_TRENDLINE and stay short if price is trading below HT_TRENDLINE.
tb.HT_TRENDLINE(np_real_data[-200:])[np.isnan(tb.HT_TRENDLINE(np_real_data[-500:]))].__len__()
Out[154]:
In [136]:
np_real_data / tb.HT_TRENDLINE(np_real_data)
Out[136]:
In [165]:
n
Out[165]:
In [43]:
dataClean.Close.values
Out[43]:
In [27]:
dataClean[dataClean['OpenRatioModi'] > 0.023]['nextpnl'].hist()
Out[27]:
In [28]:
dataClean['pnl'].mean()
Out[28]:
In [29]:
# check the monotonity of pnl filtered by openRatio indicator in bullish market
variation = []
bullpnlMeanList = []
bullpnlMedianList = []
x_axis = np.linspace(0,0.03,50)
for i in x_axis:
variation.append(i)
bullpnlMeanList.append(dataClean[dataClean['OpenRatioModi'] >= i]['pnl'].mean())
bullpnlMedianList.append(dataClean[dataClean['OpenRatioModi'] >= i]['pnl'].shape[0] / dataClean.shape[0])
fig, (ax1,ax2) = plt.subplots(nrows=2,sharex=True,figsize=(16,10))
ax1.plot(x_axis,bullpnlMeanList,'blue',label='pnlMean')
ax2.plot(x_axis,bullpnlMedianList,'red',label='pnlMedian')
ax1.legend()
ax2.legend()
Out[29]:
In [ ]:
新指标做多阈值大概为0.02(0.85左右的分位数)，做空为0.024(0.8的分位数),也可取（0.85的分位数0.0275）
In [30]:
np.sqrt(np.array([1,2,3,4]))/np.sqrt(np.array([1,2,3,4])).sum()
Out[30]:
In [31]:
dataClean['OpenRatio'].hist()
Out[31]:
In [32]:
dataClean['OpenRatioModi'].hist()
Out[32]:
In [33]:
a,b,c = 10000,3000,7000
print np.sqrt(a) / (np.sqrt(a)+np.sqrt(b)+np.sqrt(c))
print np.sqrt(b) / (np.sqrt(a)+np.sqrt(b)+np.sqrt(c))
print np.sqrt(c) / (np.sqrt(a)+np.sqrt(b)+np.sqrt(c))
In [34]:
print 3/(1+2+3)
print 1/(1+2+3)
print 2/(1+2+3)
In [35]:
print np.log10(a) / (np.log10(a)+np.log10(b)+np.log10(c))
print np.log10(b) / (np.log10(a)+np.log10(b)+np.log10(c))
print np.log10(c) / (np.log10(a)+np.log10(b)+np.log10(c))
In [36]:
print np.log(a) / (np.log(a)+np.log(b)+np.log(c))
print np.log(b) / (np.log(a)+np.log(b)+np.log(c))
print np.log(c) / (np.log(a)+np.log(b)+np.log(c))
In [ ]:
In [ ]:
In [37]:
#c = data.loc[startTime:].groupby(pd.TimeGrouper(freq='5Min'))
In [38]:
#nlarge = c['TotalVolume'].nlargest()
In [39]:
#nlarge.iloc[:50]
In [40]:
# Rescale openratio by the rank of Volume corresponding trading minute. e.g.Weighing the Open ratio by the order of total volume.
a = c['TotalVolume'].rank()/15 * data['OpenRatio']
dataslice = data.loc[a.index]
dataslice['ModiOR'] = a
dataslice.groupby(pd.TimeGrouper(freq='5Min'))['OpenRatio'].mean().dropna().describe()
In [ ]:
dataslice.groupby(pd.TimeGrouper(freq='5Min'))['OpenRatio'].mean().dropna().hist()
In [ ]:
dataModi = dataslice.groupby(pd.TimeGrouper(freq='5Min'))['OpenRatio'].mean().dropna()
dataModi.quantile(0.85)
In [ ]:
def resample_data(data,period='5Min'):
# 将一分钟线聚合为5分钟线
databy5min = pd.DataFrame()
groupgenerator = data.groupby(pd.TimeGrouper(freq=period))
databy5min['Open'] = groupgenerator['Open'].first()
databy5min['High'] = groupgenerator['High'].max()
databy5min['Low'] = groupgenerator['Low'].min()
databy5min['Close'] = groupgenerator['Close'].last()
databy5min['TotalVolume'] = groupgenerator['TotalVolume'].sum()
databy5min['OpenInterest'] = groupgenerator['OpenInterest'].last()
databy5min.dropna(axis=0,inplace = True)
return databy5min
In [ ]:
dfResampled = resample_data(data)
In [ ]:
dfResampled['OpenRatio'] = (dfResampled['OpenInterest'] - dfResampled['OpenInterest'].shift(1)) / dfResampled['TotalVolume']
dfResampled['WeightedOpenRatio'] = (dfResampled['OpenInterest'] - dfResampled['OpenInterest'].shift(1)) / np.sqrt(dfResampled['TotalVolume'])
dfResampled['pnl'] = dfResampled['Close'].pct_change()
In [ ]:
sampleSurge= dfResampled.loc[startTimeSurge:endTimeSurge]
sampleSurge['nextpnl'] = sampleSurge['pnl'].shift(-1)
In [ ]:
sampleSurge
In [ ]:
samplePlunge = dfResampled.loc[startTimePlunge:endTimePlunge]
samplePlunge['nextpnl'] = samplePlunge['pnl'].shift(-1)
In [ ]:
samplePlunge['OpenRatio'].hist()
In [ ]:
sampleSurge['nextpnl'].describe()
In [ ]:
sampleSurge['OpenRatio'].hist()
In [ ]:
sampleSurge[sampleSurge['OpenRatio'] > .15]['nextpnl'].hist()
In [ ]:
sampleSurge[sampleSurge['OpenRatio'] > 0.03]['nextpnl'].describe()
In [ ]:
In [ ]:
# check the monotonity of pnl filtered by openRatio indicator in bullish market
variation = []
bullpnlMeanList = []
bullpnlMedianList = []
for i in np.linspace(0,0.2,40):
variation.append(i)
bullpnlMeanList.append(sampleSurge[sampleSurge['OpenRatio'] >= i]['pnl'].mean())
bullpnlMedianList.append(sampleSurge[sampleSurge['OpenRatio'] >= i].shape[0] / sampleSurge.shape[0])
fig, (ax1,ax2) = plt.subplots(nrows=2,sharex=True,figsize=(16,10))
ax1.plot(np.linspace(0,0.2,40),bullpnlMeanList,'blue',label='pnlMean')
ax2.plot(np.linspace(0,0.2,40),bullpnlMedianList,'red',label='pnlMedian')
ax1.legend()
ax2.legend()
In [ ]:
# check the monotonity of pnl filtered by openRatio indicator in bearrih market
variation = []
bearpnlMeanList = []
bearpnlMedianList = []
for i in np.linspace(0,0.2,40):
variation.append(i)
bearpnlMeanList.append(samplePlunge[samplePlunge['OpenRatio'] >= i]['nextpnl'].mean())
bearpnlMedianList.append(samplePlunge[samplePlunge['OpenRatio'] >= i].shape[0] / samplePlunge.shape[0])
fig, (ax1,ax2) = plt.subplots(nrows=2,sharex=True,figsize=(16,10))
ax1.plot(np.linspace(0,0.2,40),bearpnlMeanList,'blue',label='pnlMean')
ax2.plot(np.linspace(0,0.2,40),bearpnlMedianList,'red',label='pnlMedian')
ax1.legend()
ax2.legend()
In [ ]:
In [ ]:
sampleSurge[sampleSurge['OpenRatio'] >= 0.1]['nextpnl'].hist()
In [ ]:
pnlMedianList
In [ ]:
dfResampled['OpenRatio'].loc[startTime:].hist()
In [ ]:
dfResampled['WeightedOpenRatio'].quantile(0.996)
In [ ]:
dfResampled[dfResampled['OpenRatio'] > 0].mean()
In [ ]:
dfResampled['OpenRatio'].tail(500).plot()
In [ ]:
dfResampled['Close'].tail(30).plot()
In [ ]:
dfResampled[['pnl','OpenRatio']].tail(50)
In [ ]:
len(dfResampled[dfResampled['OpenRatio'] < 0.15]) / len(dfResampled)
In [ ]:
len(dfResampled[dfResampled['OpenRatio'] < 0]) / len(dfResampled)
In [ ]:
np.sqrt(0.33*225/5)
In [ ]:
datatrain = dfResampled.loc[startTrainTime:endTimeTrain]
In [ ]:
startTime
In [ ]:
datatrain['EMAOpenRatio'] = talib.EMA(datatrain['OpenRatio'].values, timeperiod=4)
datatrain['EMAOpenRatio'].describe()
In [ ]:
datatrain['OpenRatio'].describe()
In [ ]:
datatrain['EMAOpenRatio'].quantile(0.92)
In [ ]:
datatrain['OpenRatio'].quantile(0.9)
In [ ]:
# show size distribution(quantile plot)
fig = plt.figure(figsize=(16,10))
# Add a subplot
ax = fig.add_subplot(111)
datatrain['EMAOpenRatio'].hist(alpha =0.8)
datatrain['OpenRatio'].hist(alpha = 0.3)
ax.set_title(ax.get_title(),alpha=0.7, fontsize=25)
In [ ]:
datatrain['EMAOpenRatio'].hist()
datatrain['OpenRatio'].hist(alpha = 0.3)
In [ ]:
datatrain['OpenRatio'].describe()
In [ ]:
In [ ]:
dfResampled['Close'].plot(figsize=(18,12))
In [ ]:
shapeNum = 20
In [ ]:
svdindicator = getSVD(datatrain['Close'].values[-shapeNum:], shapeNum, svdShort)
In [ ]:
In [ ]:
svdindicator1 = getSVD(datatrain['Close'].values[-shapeNum:], shapeNum, svdLong)
svdindicator1
In [ ]:
fig, (ax1, ax2) = plt.subplots(nrows=2, sharex=True,figsize=(16,10))
ax1.plot(datatrain.index[-shapeNum:],datatrain['Close'].values[-shapeNum:],'blue',label='Close')
ax2.plot(datatrain.index[-shapeNum:],svdindicator,'green',)
ax2.plot(datatrain.index[-shapeNum:],svdindicator1,'black')
#ax1.title('SVD')
In [ ]:
fig, (ax1, ax2) = plt.subplots(nrows=2, sharex=True,figsize=(22,16))
ax1.plot(datatrain['Close'].values[-shapeNum:],'blue',label='Close')
ax2.plot(svdindicator,'red',label='svd-'+str(svdShort))
ax2.plot(svdindicator1,'yellow',label='svd-'+str(svdLong))
#ax1.title('SVD')
ax1.legend()
ax2.legend()
In [ ]:
datatrain['Close'].plot(figsize=(18,12))
In [ ]:
#kvol.median()
In [ ]:
kvol = (datatrain['Close'] - datatrain['Close'].shift(1)).rolling(min_periods=60,window=60,center=False).std()
In [ ]:
kvol.iloc[-100:].plot(figsize=(18,12))
((7*datatrain['Close']/datatrain['Close'].iloc[0]).iloc[-100:]).plot(sharex=True)
In [ ]:
datatrain['Close'].iloc[-500:].plot(figsize=(18,12))
In [ ]:
datatrain['Impact'] = (datatrain['Close'] - datatrain['Open']) / datatrain['TotalVolume']
In [ ]:
datatrain['Impact'].iloc[-55:].plot(figsize=(18,12))
In [ ]:
In [ ]:
datatrain['TotalVolume'].iloc[-55:].mean()
In [ ]:
datatrain[datatrain['Impact']>0]
In [ ]:
filtered = datatrain[(datatrain['TotalVolume']<datatrain['TotalVolume'].quantile(0.2)) & (datatrain['Impact']>((datatrain[datatrain['Impact']>0]['Impact']).quantile(0.7)))]
In [ ]:
filtered
In [ ]:
datatrain.iloc[:2]
In [ ]:
filtered['Close'].plot(figsize=(18,10))
In [ ]:
trackBackPeriod = 50
code = 'rb000'
initialNum = 100
In [ ]:
In [ ]:
futurebuyList = []
for i in range(initialNum,len(datatrain)-1):
svdshortArray = getSVD(datatrain['Close'].values[i+1-trackBackPeriod:i+1], trackBackPeriod, svdShort)
svdlongArray = getSVD(datatrain['Close'].values[i+1-trackBackPeriod:i+1], trackBackPeriod, svdLong)
if svdshortArray[-2] < svdlongArray[-2] and svdshortArray[-1] > svdlongArray[-1]:
dfslice = datatrain.iloc[i+1]
amount = 1
price = dfslice['Open']
cost = price * amount
direction = 1
futurebuyList.append([dfslice.name,code,price,amount,cost,direction])
In [ ]:
futurebuyList
In [ ]:
In [105]:
filefullname = 'C:/Users/LZJF_02/Desktop/temp.xlsx'
In [106]:
xls = pd.ExcelFile(filefullname)
df1 = xls.parse('file',header=0,index_col=0,parse_dates=True)
In [116]:
df = df1.dropna(how='all',axis=1)
df=df.dropna(axis=0)
df.columns = ['open','high','low',u'贵州茅台','amount','volume','hs300',u'白酒指数']
df = df[[u'贵州茅台',u'白酒指数']]
df=df.dropna(axis=0)
df.index.name = 'time'
In [117]:
df
Out[117]:
In [118]:
df.index = pd.to_datetime(df.index)
In [119]:
df.index
Out[119]:
In [120]:
endOfWeekList = sorted(list(set(df.iloc[df.resample('W').size().cumsum().sub(1)].index)))
In [121]:
vol = df.loc[endOfWeekList].pct_change().rolling(window= 20, min_periods=20).std().round(3)
In [122]:
vol
Out[122]:
In [123]:
# make sure that matplotib and seaborn can show Chinese
import matplotlib as mpl
mpl.rcParams['font.sans-serif'] = ['SimHei']
mpl.rcParams['font.serif'] = ['SimHei']
sns.set_style("darkgrid",{"font.sans-serif":['simhei', 'Arial']})
In [124]:
# show size distribution(quantile plot)
fig = plt.figure(figsize=(16,10))
# Add a subplot
ax = fig.add_subplot(111)
vol.plot(figsize=(22,14),ax=ax,title ='Volatility Difference',fontsize =13)
ax.set_title(ax.get_title(),alpha=0.7, fontsize=25)
Out[124]: