def computeRunningMinMaxSignals(df, ema_feats_df, spike_5s, norm):
max_to_here = pd.expanding_max(df['microprice_ema_200ms_ticks'])
min_to_here = pd.expanding_min(df['microprice_ema_200ms_ticks'])
if spike_5s >= 0:
max_disl = max_to_here
else:
max_disl = min_to_here
ema_feats_df['from_max_disl_ema_200ms_ticks'] = computeRelativeDislocation(df['microprice_ema_200ms_ticks'], max_disl, max_disl)
ema_feats_df['from_max_disl_ema_2000ms_ticks'] = computeRelativeDislocation(df['microprice_ema_2000ms_ticks'], max_disl, max_disl)
ema_feats_df['from_max_disl_ema_10000ms_ticks'] = computeRelativeDislocation(df['microprice_ema_10000ms_ticks'], max_disl, max_disl)
rev_from_max_disl_to_here = df['microprice_ema_200ms_ticks'] - max_disl
max_rev_to_here = pd.expanding_max(rev_from_max_disl_to_here)
min_rev_to_here = pd.expanding_min(rev_from_max_disl_to_here)
if spike_5s >= 0:
max_rev = max_disl + min_rev_to_here
else:
max_rev = max_disl + max_rev_to_here
ema_feats_df['from_max_rev_ema_200ms_ticks'] = computeRelativeDislocation(df['microprice_ema_200ms_ticks'], max_rev, max_rev)
ema_feats_df['from_max_rev_ema_2000ms_ticks'] = computeRelativeDislocation(df['microprice_ema_2000ms_ticks'], max_rev, max_rev)
ema_feats_df['from_max_rev_ema_10000ms_ticks'] = computeRelativeDislocation(df['microprice_ema_10000ms_ticks'], max_rev, max_rev)
return (ema_feats_df, df)
def update_price(self, date, sid):
try:
p = pd.Series(self.collection.find_one({'sid': sid, 'date': date, 'dname': 'price'})['dvalue'])
except:
self.logger.warning('No price found for {} on {}', sid, date)
return
p.index = [datetime.strptime(date, '%Y%m%d')+timedelta(milliseconds=int(s)) for s in p.index]
df5 = p.resample('5min', 'ohlc', label='right', closed='right')
df5.columns = ['into', 'inth', 'intl', 'intc']
df5['lstl'] = pd.expanding_min(p).resample('5min', 'last', label='right', closed='right')
df5['lsth'] = pd.expanding_max(p).resample('5min', 'last', label='right', closed='right')
df5['vlty'] = p.resample('5min', lambda x: x.std(), label='right', closed='right')
df5['lstp'] = p.resample('5min', 'last', label='right', closed='right')
df5['tvwp'] = p.resample('5min', 'mean', label='right', closed='right')
df5.index = [dt.strftime('%H%M%S') for dt in df5.index]
df5 = df5.ix[times_5min]
for dname, ser in df5.iteritems():
self.db.IF_5min.update({'sid': sid, 'date': date, 'dname': dname}, {'$set': {'dvalue': ser.to_dict()}}, upsert=True)
df1 = p.resample('1min', 'ohlc', label='right', closed='right')
df1.columns = ['into', 'inth', 'intl', 'intc']
df1['lstl'] = pd.expanding_min(p).resample('1min', 'last', label='right', closed='right')
df1['lstl'] = pd.expanding_max(p).resample('1min', 'last', label='right', closed='right')
df1['vlty'] = p.resample('1min', lambda x: x.std(), label='right', closed='right')
df1['lstp'] = p.resample('1min', 'last', label='right', closed='right')
df1.index = [dt.strftime('%H%M%S') for dt in df1.index]
df1 = df1.ix[times_1min]
for dname, ser in df1.iteritems():
self.db.IF_1min.update({'sid': sid, 'date': date, 'dname': dname}, {'$set': {'dvalue': ser.to_dict()}}, upsert=True)
def indicator_KDJ(stock_data): #KDJ指标计算函数
# 计算KDJ指标
low_list = pd.rolling_min(stock_data['low'], 9) #9天为一个周期，但前8个值为NaN
low_list.fillna(value=pd.expanding_min(stock_data['low']),
inplace=True) #将NaN用累积窗口计算的最小值代替
high_list = pd.rolling_max(stock_data['high'], 9)
high_list.fillna(value=pd.expanding_max(stock_data['high']), inplace=True)
rsv = (stock_data['close'] - low_list) / (high_list - low_list) * 100
stock_data['KDJ_K'] = pd.ewma(rsv, com=2, adjust=False)
stock_data['KDJ_D'] = pd.ewma(stock_data['KDJ_K'], com=2, adjust=False)
stock_data['KDJ_J'] = 3 * stock_data['KDJ_K'] - 2 * stock_data['KDJ_D']
# 计算KDJ指标金叉、死叉情况
stock_data['KDJ_金叉死叉'] = ''
kdj_position = stock_data['KDJ_K'] > stock_data['KDJ_D']
stock_data.loc[kdj_position[(kdj_position == True)
& (kdj_position.shift() == False)].index,
'KDJ_金叉死叉'] = '金叉' #前一天K<D,当天K>D
stock_data.loc[kdj_position[(kdj_position == False)
& (kdj_position.shift() == True)].index,
'KDJ_金叉死叉'] = '死叉'
# 通过复权价格计算接下来几个交易日的收益率
for n in [1, 2, 3, 5, 10, 20]:
stock_data['接下来' + str(n) + '个交易日涨跌幅'] = stock_data['close'].shift(
-1 * n) / stock_data['close'] - 1.0
stock_data.dropna(how='any', inplace=True) # 删除所有有空值的数据行
# 筛选出KDJ金叉的数据，并将这些数据合并到all_stock中
stock_data = stock_data[(stock_data['KDJ_金叉死叉'] == '金叉')]
if not stock_data.empty:
return stock_data
def get_kdj(code):
stock_data = ts.get_k_data(code)
# kdj
low_list = pd.rolling_min(stock_data['low'], 9)
low_list.fillna(value=pd.expanding_min(stock_data['low']), inplace=True)
high_list = pd.rolling_max(stock_data['high'], 9)
high_list.fillna(value=pd.expanding_max(stock_data['high']), inplace=True)
rsv = (stock_data['close'] - low_list) / (high_list - low_list) * 100
stock_data['kdj_k'] = pd.ewma(rsv, com=2)
stock_data['kdj_d'] = pd.ewma(stock_data['kdj_k'], com=2)
stock_data['kdj_j'] = 3 * stock_data['kdj_k'] - 2 * stock_data['kdj_d']
# 用今天的j值和昨天比较
kdj_j = stock_data['kdj_j']
yesterdayJ = kdj_j[kdj_j.size - 2]
todayJ = kdj_j[kdj_j.size - 1]
kdj_k = stock_data['kdj_k']
todayK = kdj_k[kdj_k.size - 1]
# 如果今天的j值大于昨天的j值才继续后面的逻辑
if (todayJ > yesterdayJ and todayK < float(20)):
# 计算价格5日百分比
stock_data = stock_data[stock_data.date > str(dc.get_the_day_before_today(1))]
stock_data['kdj_ok'] = 1
else:
stock_data = stock_data[stock_data.date > str(dc.get_the_day_before_today(1))]
stock_data['kdj_ok'] = 0
return stock_data
def KDJ(date, N=9, M1=3, M2=3):
low_list = pd.rolling_min(date[low], N)
low_list.fillna(value=pd.expanding_min(date[low]), inplace=True)
high_list = pd.rolling_max(date[high], N)
high_list.fillna(value=pd.expanding_max(date[high]), inplace=True)
rsv = (date['close'] - low_list) / (high_list - low_list) * 100
date['KDJ_K'] = pd.ewma(rsv, com=2)
date['KDJ_D'] = pd.ewma(date['KDJ_K'], com=2)
date['KDJ_J'] = 3 * date['KDJ_K'] - 2 * date['KDJ_D']
def kdj(data,date,m1,m2): data_use=data[['high','low','open','close']] data['lown'] = pd.rolling_min(data_use['low'], date) data.lown.fillna(value=pd.expanding_min(data_use['low']), inplace=True) data['highn'] = pd.rolling_max(data_use['high'], date) data.highn.fillna(value=pd.expanding_max(data_use['high']), inplace=True) data['rsv']=(data['close'] - data['lown']) / (data['highn'] - data['lown']) * 100 data['kdj_k'] = pd.ewma(data['rsv'], m1) data['kdj_d'] = pd.ewma(data['kdj_k'], m2) data['kdj_j'] = 3 * data['kdj_k'] - 2 * data['kdj_d']
def kdj(data,date,m1,m2): data_use=data[['high','low','open','close']] data['lown'] = pd.rolling_min(data_use['low'], date) data.lown.fillna(value=pd.expanding_min(data_use['low']), inplace=True) data['highn'] = pd.rolling_max(data_use['high'], date) data.highn.fillna(value=pd.expanding_max(data_use['high']), inplace=True) data['rsv']=(data['close'] - data['lown']) / (data['highn'] - data['lown']) * 100 data['kdj_k'] = pd.ewma(data['rsv'], m1) data['kdj_d'] = pd.ewma(data['kdj_k'], m2) data['kdj_j'] = 3 * data['kdj_k'] - 2 * data['kdj_d']
def analysis_kdjv2(code):
filename='./stockdata/data/last3year/'+code+'.csv'
stock_dataT = pd.read_csv(filename, parse_dates=['date'])
#stock_dataT.sort('date', inplace=True)
stock_data=stock_dataT.loc[:,('date', 'high', 'low', 'close', 'p_change')]
# 计算KDJ指标
stock_data['low_list'] = pd.rolling_min(stock_data['low'], 9)
stock_data['low_list'].fillna(value=pd.expanding_min(stock_data['low']), inplace=True)
stock_data['high_list'] = pd.rolling_max(stock_data['high'], 9)
stock_data['high_list'].fillna(value=pd.expanding_max(stock_data['high']), inplace=True)
stock_data['rsv'] = (stock_data['close'] - stock_data['low_list']) / (stock_data['high_list'] - stock_data['low_list']) * 100
#stock_data['rsv']=(stock_data['close'] - low_list) / (high_list - low_list) * 100
stock_data['KDJ_K'] = pd.ewma(stock_data['rsv'], com=3)
stock_data['KDJ_D'] = pd.ewma(stock_data['KDJ_K'], com=3)
stock_data['KDJ_J'] = 3 * stock_data['KDJ_K'] - 2 * stock_data['KDJ_D']
# 计算KDJ指标金叉、死叉情况
###通常就敏感性而言，J值最强，K值次之，D值最慢，而就安全性而言，J值最差，K值次之，D值最稳
##金叉用1表示，死叉用0表示
buyi=stock_data[(stock_data['KDJ_K'] > stock_data['KDJ_D'])&(stock_data['KDJ_K'].shift(1) < stock_data['KDJ_D'].shift(1))].index
stock_data.loc[buyi,'Signal'] = 1
selli=stock_data[(stock_data['KDJ_K'] < stock_data['KDJ_D'])&(stock_data['KDJ_K'].shift(1) > stock_data['KDJ_D'].shift(1))].index
stock_data.loc[selli,'Signal'] = 0
#kdj_position = stock_data['KDJ_K'] > stock_data['KDJ_D']
#stock_data.loc[kdj_position[(kdj_position == True) & (kdj_position.shift() == False)].index, 'KDJ_BS'] = 1
#stock_data.loc[kdj_position[(kdj_position == False) & (kdj_position.shift() == True)].index, 'KDJ_BS'] = 0
stock_data['position']=stock_data['Signal'].shift(1)
stock_data['position'].fillna(method='ffill', inplace=True)
#当仓位为1时，已当天的开盘价买入股票，当仓位为0时，以收盘价卖出该股份。计算从数据期内的收益
stock_data['Cash_index'] = ((stock_data['p_change']/100) * stock_data['position'] + 1.0).cumprod()
initial_idx = 1
#initial_idx = stock_data.iloc[0]['close'] / (1 + (stock_data.iloc[0]['p_change']/100))
stock_data['Cash_index'] *= initial_idx
print 'The KDJ Backwards methon Signal:'
Make_decision(stock_data)
# 通过复权价格计算接下来几个交易日的收益率
for n in [1, 2, 3, 5, 10, 20]:
stock_data['CP_next_'+str(n)+'_days'] =(stock_data['close'].shift(-1*n) / stock_data['close'] - 1.0)*100
#stock_data.dropna(how='any', inplace=True)# 删除所有有空值的数据行
# ========== 将算好的数据输出到csv文件 - 注意：这里请填写输出文件在您电脑中的路径
##统计出现买点时点的数据
dd=stock_data[stock_data['Signal']==1]
print_return_next_n_day(dd)
codedir='./output/A/'+code+os.sep
if not os.path.exists(codedir):
os.mkdir(codedir)
# ==========计算每年指数的收益以及海龟交易法则的收益
stock_data['p_change_KDJV2'] = (stock_data['p_change']) * stock_data['position']
year_rtn = stock_data.set_index('date')[['p_change', 'p_change_KDJV2']].\
resample('A', how=lambda x: (x/100+1.0).prod() - 1.0) * 100
year_rtn.to_csv(codedir+'kdjv2_year.csv', encoding='gbk')
stock_data.to_csv(codedir+'kdjv2.csv',encoding='gbk',index=False)
stock_data.tail(20).to_csv(codedir+'kdjv2_Signal.csv',encoding='gbk',index=False)
print 'the share %s trading sign for KDJV2:'%code
print stock_data.tail(5)
return
def kdj(stock):
low_list = pd.rolling_min(stock.low, 9)
low_list.fillna(value=pd.expanding_min(stock.low), inplace=True)
high_list = pd.rolling_max(stock.high, 9)
high_list.fillna(value=pd.expanding_max(stock.high), inplace=True)
rsv = (stock.close - low_list) / (high_list - low_list) * 100
k = pd.ewma(rsv, com=2)
d = pd.ewma(k, com=2)
j = 3 * k[2:] - 2 * d
return k, d, j
def kdj(stock):
low_list = pd.rolling_min(stock.low, 9)
low_list.fillna(value = pd.expanding_min(stock.low), inplace = True)
high_list = pd.rolling_max(stock.high, 9)
high_list.fillna(value = pd.expanding_max(stock.high), inplace = True)
rsv = (stock.close - low_list) / (high_list - low_list) * 100
k = pd.ewma(rsv, com =2)
d = pd.ewma(k, com =2)
j = 3 * k[2:] - 2 *d
return k, d, j
def get_kdj(stock_data):
# kdj计算
low_list = pd.rolling_min(stock_data['low'], 9)
low_list.fillna(value=pd.expanding_min(stock_data['low']), inplace=True)
high_list = pd.rolling_max(stock_data['high'], 9)
high_list.fillna(value=pd.expanding_max(stock_data['high']), inplace=True)
rsv = (stock_data['close'] - low_list) / (high_list - low_list) * 100
# 增加kdj数据到 stock_data中
stock_data['kdj_k'] = round(pd.ewma(rsv, com=2), 2)
stock_data['kdj_d'] = round(pd.ewma(stock_data['kdj_k'], com=2), 2)
stock_data['kdj_j'] = round(
3 * stock_data['kdj_k'] - 2 * stock_data['kdj_d'], 2)
return stock_data
def calKDJ(data, N=0, M=0):
if N == 0:
N = 9
if M == 0:
M = 2
low_list = pd.rolling_min(data['low'], N)
low_list.fillna(value=pd.expanding_min(data['low']), inplace=True)
high_list = pd.rolling_max(data['high'], N)
high_list.fillna(value=pd.expanding_max(data['high']), inplace=True)
rsv = (data['close'] - low_list) / (high_list - low_list) * 100
KDJ_K = pd.ewma(rsv, com=M)
KDJ_D = pd.ewma(KDJ_K, com=M)
KDJ_J = 3 * KDJ_K - 2 * KDJ_D
#kdjdata.fillna(0, inplace=True)
return low_list, high_list, rsv, KDJ_K, KDJ_D, KDJ_J
def analyse(self):
# Logger.log(logging.INFO, "Analyse Strategy", {"scope":__name__, "Rule 1":self._rule1, "Rule 2":self._rule2, "Rule 3":self._rule3, "Type":self._type})
connection = sqlite3.connect(pyswing.database.pySwingDatabase)
query = self.analyseStrategySql % (self._rule1, self._rule2, self._rule3, self._exit, self._type)
self._strategyData = read_sql_query(query, connection, 'Date')
self._strategyData['ExitValueAfterCosts'] = self._strategyData['ExitValue'] - 0.2
connection.close()
exitValueDataFrame = self._strategyData.ix[:,'ExitValueAfterCosts']
mean = exitValueDataFrame.mean()
median = exitValueDataFrame.median()
sum = exitValueDataFrame.sum()
count = exitValueDataFrame.count()
tradesPerYear = count / 10
sharpeRatio = sqrt(tradesPerYear) * exitValueDataFrame.mean() / exitValueDataFrame.std()
self._strategyData["Sum"] = expanding_sum(exitValueDataFrame)
self._strategyData["Max"] = expanding_max(self._strategyData["Sum"])
self._strategyData["Min"] = expanding_min(self._strategyData["Sum"])
self._strategyData["DD"] = self._strategyData["Max"] - self._strategyData["Min"]
runningSum = expanding_sum(exitValueDataFrame)
max2here = expanding_max(runningSum)
dd2here = runningSum - max2here
drawDown = dd2here.min()
Logger.log(logging.INFO, "Analysing Strategy", {"scope":__name__, "Rule 1":self._rule1, "Rule 2":self._rule2, "Rule 3":self._rule3, "Exit":self._exit, "Type":self._type, "Mean":str(mean), "Median":str(median), "Sum":str(sum), "Count":str(count), "SharpeRatio":str(sharpeRatio), "DrawDown":str(drawDown)})
connection = sqlite3.connect(pyswing.database.pySwingDatabase)
c = connection.cursor()
deleteSql = self.deleteStrategySql % (pyswing.globals.pySwingStrategy, self._rule1, self._rule2, self._rule3, self._exit, self._type)
c.executescript(deleteSql)
connection.commit()
insertSql = self.insertStrategySql % (pyswing.globals.pySwingStrategy, self._rule1, self._rule2, self._rule3, self._exit, self._type, str(mean), str(median), str(sum), str(count), str(sharpeRatio), str(drawDown))
c.executescript(insertSql)
connection.commit()
c.close()
connection.close()
def get_kdj_history(code):
stock_data = ts.get_k_data(code)
# kdj
low_list = pd.rolling_min(stock_data['low'], 9)
low_list.fillna(value=pd.expanding_min(stock_data['low']), inplace=True)
high_list = pd.rolling_max(stock_data['high'], 9)
high_list.fillna(value=pd.expanding_max(stock_data['high']), inplace=True)
rsv = (stock_data['close'] - low_list) / (high_list - low_list) * 100
stock_data['kdj_k'] = round(pd.ewma(rsv, com=2), 2)
stock_data['kdj_d'] = round(pd.ewma(stock_data['kdj_k'], com=2), 2)
stock_data['kdj_j'] = round(3 * stock_data['kdj_k'] - 2 * stock_data['kdj_d'], 2)
# 用今天的j值和昨天比较
kdj_j = stock_data['kdj_j']
if (kdj_j.size < 6):
stock_data = stock_data.tail(1)
stock_data['kdj_k'] = 0
stock_data['kdj_ok'] = 0
return stock_data
yesterdayJ = kdj_j[kdj_j.size - 6]
todayJ = kdj_j[kdj_j.size - 5]
kdj_k = stock_data['kdj_k']
todayK = kdj_k[kdj_k.size - 5]
# 如果今天的j值大于昨天的j值才继续后面的逻辑
if (todayJ > yesterdayJ and todayK < float(20)):
# 计算价格5日百分比
stock_data_copy = stock_data[:]
stock_data_copy = stock_data_copy.tail(5)
stock_data_copy['indexNum'] = [1, 2, 3, 4, 5]
stock_data_copy = stock_data_copy.sort(columns='high')
stock_data_copy = stock_data_copy.tail(1)
maxValue = stock_data_copy.high.values
maxDate = stock_data_copy.date.values
stock_data = stock_data.tail(5)
stock_data = stock_data.head(1)
stock_data['kdj_ok'] = 1
highPercent = maxValue / stock_data.close.values[0]
stock_data['highPercent'] = (round(highPercent, 3) * 100) - 100
stock_data['highDate'] = maxDate
stock_data['highDays'] = stock_data_copy.indexNum.values
else:
stock_data = stock_data.tail(1)
stock_data['kdj_ok'] = 0
return stock_data
def expanding_smoother(self, data, stype='rolling_mean', min_periods=None, freq=None):
"""
Perform a expanding smooting on the data for a complete help refer to http://pandas.pydata.org/pandas-docs/dev/computation.html
:param data: pandas dataframe input data
:param stype: soothing type
:param min_periods: periods
:param freq: frequence
smoothing types:
expanding_count Number of non-null observations
expanding_sum Sum of values
expanding_mean Mean of values
expanding_median Arithmetic median of values
expanding_min Minimum
expanding_max Maximum
expandingg_std Unbiased standard deviation
expanding_var Unbiased variance
expanding_skew Unbiased skewness (3rd moment)
expanding_kurt Unbiased kurtosis (4th moment)
"""
if stype == 'count':
newy = pd.expanding_count(data, min_periods=min_periods, freq=freq)
if stype == 'sum':
newy = pd.expanding_sum(data, min_periods=min_periods, freq=freq)
if stype == 'mean':
newy = pd.expanding_mean(data, min_periods=min_periods, freq=freq)
if stype == 'median':
newy = pd.expanding_median(data, min_periods=min_periods, freq=freq)
if stype == 'min':
newy = pd.expanding_min(data, min_periods=min_periods, freq=freq)
if stype == 'max':
newy = pd.expanding_max(data, min_periods=min_periods, freq=freq)
if stype == 'std':
newy = pd.expanding_std(data, min_periods=min_periods, freq=freq)
if stype == 'var':
newy = pd.expanding_var(data, min_periods=min_periods, freq=freq)
if stype == 'skew':
newy = pd.expanding_skew(data, min_periods=min_periods, freq=freq)
if stype == 'kurt':
newy = pd.expanding_kurt(data, min_periods=min_periods, freq=freq)
return newy
def lm_kdj(df, n,ksgn='close'):
'''
【输入】
df, pd.dataframe格式数据源
n，时间长度
ksgn，列名，一般是：close收盘价
【输出】
df, pd.dataframe格式数据源,
增加了一栏：_{n}，输出数据
'''
lowList= pd.rolling_min(df['low'], n)
lowList.fillna(value=pd.expanding_min(df['low']), inplace=True)
highList = pd.rolling_max(df['high'], n)
highList.fillna(value=pd.expanding_max(df['high']), inplace=True)
rsv = (df[ksgn] - lowList) / (highList - lowList) * 100
df['k'] = pd.ewma(rsv,com=2)
df['d'] = pd.ewma(df['k'],com=2)
df['j'] = 3.0 * df['k'] - 2.0 * df['d']
#print('n df',len(df))
return df
def computeRunningMinMaxSignals(self, df, start_dt, spike_ticks, spike_5s_pred):
# print start_dt
pre_event_snapshot_dt = start_dt - timedelta(seconds=2.5)
pre_event_snapshot_loc = df.index.get_loc(pre_event_snapshot_dt)
# print df.ix[self.cur_loc, 'time'], len(df.ix[pre_event_snapshot_loc : self.cur_loc+1, 'microprice_ema_200ms'])
max_to_here = pd.expanding_max(df.ix[pre_event_snapshot_loc : self.cur_loc+1, 'microprice_ema_200ms'])[-1]
min_to_here = pd.expanding_min(df.ix[pre_event_snapshot_loc : self.cur_loc+1, 'microprice_ema_200ms'])[-1]
max_to_here_ticks = self.priceTicks(max_to_here)
min_to_here_ticks = self.priceTicks(min_to_here)
if (spike_ticks + spike_5s_pred)/2.0 >= 0:
max_disl = max_to_here_ticks
else:
max_disl = min_to_here_ticks
#df.ix[self.cur_loc, 'from_max_disl_ema_200ms_ticks'] = computeRelativeDislocation(df.ix[self.cur_loc, 'microprice_ema_200ms_ticks'], max_disl)
df.ix[self.cur_loc, 'max_disl_ema_200ms_ticks'] = max_disl
#df[self.cur_loc]['max_disl_ema_200ms_ticks'] = max_disl
return df
def indicator_KDJ(stock_data):#KDJ指标计算函数
# 计算KDJ指标
low_list = pd.rolling_min(stock_data['low'], 9) #9天为一个周期，但前8个值为NaN
low_list.fillna(value=pd.expanding_min(stock_data['low']), inplace=True) #将NaN用累积窗口计算的最小值代替
high_list = pd.rolling_max(stock_data['high'], 9)
high_list.fillna(value=pd.expanding_max(stock_data['high']), inplace=True)
rsv = (stock_data['close'] - low_list) / (high_list - low_list) * 100
stock_data['KDJ_K'] = pd.ewma(rsv, com=2, adjust=False)
stock_data['KDJ_D'] = pd.ewma(stock_data['KDJ_K'], com=2, adjust=False)
stock_data['KDJ_J'] = 3 * stock_data['KDJ_K'] - 2 * stock_data['KDJ_D']
# 计算KDJ指标金叉、死叉情况
stock_data['KDJ_金叉死叉'] = ''
kdj_position = stock_data['KDJ_K'] > stock_data['KDJ_D']
stock_data.loc[kdj_position[(kdj_position == True) & (kdj_position.shift() == False)].index, 'KDJ_金叉死叉'] = '金叉' #前一天K<D,当天K>D
stock_data.loc[kdj_position[(kdj_position == False) & (kdj_position.shift() == True)].index, 'KDJ_金叉死叉'] = '死叉'
# 通过复权价格计算接下来几个交易日的收益率
for n in [1, 2, 3, 5, 10, 20]:
stock_data['接下来'+str(n)+'个交易日涨跌幅'] = stock_data['close'].shift(-1*n) / stock_data['close'] - 1.0
stock_data.dropna(how='any', inplace=True)# 删除所有有空值的数据行
# 筛选出KDJ金叉的数据，并将这些数据合并到all_stock中
stock_data = stock_data[(stock_data['KDJ_金叉死叉'] == '金叉')]
if not stock_data.empty:
return stock_data
def plot_Ndays_Break(self, stock_df):
N1 = 42
N2 = 30
stock_df['N1_High'] = pd.rolling_max(stock_df.High,
window=N1) #计算最近N1个交易日最高价
stock_df['N1_High'] = stock_df['N1_High'].shift(1)
expan_max = pd.expanding_max(stock_df.Close)
stock_df['N1_High'].fillna(value=expan_max,
inplace=True) #目前出现过的最大值填充前N1个nan
stock_df['N2_Low'] = pd.rolling_min(stock_df.Low,
window=N2) #计算最近N2个交易日最低价
stock_df['N2_Low'] = stock_df['N2_Low'].shift(1)
expan_min = pd.expanding_min(stock_df.Close)
stock_df['N2_Low'].fillna(value=expan_min,
inplace=True) #目前出现过的最小值填充前N2个nan
dispCont_List = []
break_pd = pd.DataFrame()
for kl_index in np.arange(0, stock_df.shape[0]):
today = stock_df.ix[kl_index]
""" 收盘价超过N2最低价 卖出股票持有"""
if today['Close'] < today['N2_Low']:
break_pd = break_pd.append(today)
dispCont_List.append(
"向下突破:" + stock_df.index[kl_index].strftime('%Y-%m-%d') +
',' + str(today['Close']) + '\n') #向下突破和价格
""" 收盘价超过N1最高价 买入股票持有"""
if today['Close'] > today['N1_High']:
break_pd = break_pd.append(today)
dispCont_List.append(
"向上突破:" + stock_df.index[kl_index].strftime('%Y-%m-%d') +
',' + str(today['Close']) + '\n') #向上突破和价格
return break_pd, dispCont_List
# ==========计算海龟交易法则的买卖点
# 设定海龟交易法则的两个参数，当收盘价大于最近N1天的最高价时买入，当收盘价低于最近N2天的最低价时卖出
# 这两个参数可以自行调整大小，但是一般N1 > N2
N1 = 20
N2 = 10
# 通过rolling_max方法计算最近N1个交易日的最高价
index_data['最近N1个交易日的最高点'] = pd.rolling_max(index_data['high'], N1)
# 对于上市不足N1天的数据，取上市至今的最高价
index_data['最近N1个交易日的最高点'].fillna(value=pd.expanding_max(index_data['high']), inplace=True)
# 通过相似的方法计算最近N2个交易日的最低价
index_data['最近N2个交易日的最低点'] = pd.rolling_min(index_data['low'], N1)
index_data['最近N2个交易日的最低点'].fillna(value=pd.expanding_min(index_data['low']), inplace=True)
# 当当天的【close】> 昨天的【最近N1个交易日的最高点】时，将【收盘发出的信号】设定为1
buy_index = index_data[index_data['close'] > index_data['最近N1个交易日的最高点'].shift(1)].index
index_data.loc[buy_index, '收盘发出的信号'] = 1
# 当当天的【close】< 昨天的【最近N2个交易日的最低点】时，将【收盘发出的信号】设定为0
sell_index = index_data[index_data['close'] < index_data['最近N2个交易日的最低点'].shift(1)].index
index_data.loc[sell_index, '收盘发出的信号'] = 0
# 计算每天的仓位，当天持有上证指数时，仓位为1，当天不持有上证指数时，仓位为0
index_data['当天的仓位'] = index_data['收盘发出的信号'].shift(1)
index_data['当天的仓位'].fillna(method='ffill', inplace=True)
# 取1992年之后的数据，排出较早的数据
index_data = index_data[index_data['date'] >= pd.to_datetime('19930101')]
data1['fast_line'] = pd.rolling_mean(data1['close'], h)
data1['slow_line'] = pd.rolling_mean(data1['close'], k)
data1['fast_line'] = data1['fast_line'].fillna(
value=pd.expanding_mean(data1['close']))
data1['slow_line'] = data1['slow_line'].fillna(
value=pd.expanding_mean(data1['close']))
data1['dist_%s_%s' % (k, h)] = data1['fast_line'] - data1['slow_line']
for h in range(10, 26, 5):
data1['fast_line'] = ''
data1['slow_line'] = ''
data1['fast_line'] = pd.rolling_max(data1['high'].shift(1), h)
data1['slow_line'] = pd.rolling_min(data1['low'].shift(1), h)
data1['fast_line'] = data1['fast_line'].fillna(
value=pd.expanding_max(data1['high']))
data1['slow_line'] = data1['slow_line'].fillna(
value=pd.expanding_min(data1['low']))
data1['dist_high_%s' % h] = data1['high'] - data1['fast_line']
data1['dist_low_%s' % h] = data1['low'] - data1['slow_line']
data1 = MACD(data1, 12, 26, 9)
data2 = pd.read_csv('rb888_2017.csv', parse_dates=True, index_col='time')
data2.reset_index(inplace=True)
data2['log_return'] = np.log(data2['close'] / data2['close'].shift(1))
data2['log_return'] = data2['log_return'].fillna(0)
for h, k in [(5, 20), (15, 20), (5, 10), (5, 15), (10, 15)]:
data2['fast_line'] = ''
data2['slow_line'] = ''
data2['fast_line'] = pd.rolling_mean(data2['close'], h)
data2['slow_line'] = pd.rolling_mean(data2['close'], k)
data2['fast_line'] = data2['fast_line'].fillna(
value=pd.expanding_mean(data2['close']))
data2['slow_line'] = data2['slow_line'].fillna(
这两个参数可以自行调整大小，但是一般N1 > N2 ''' N1 = 20 N2 = 10 ''' 通过rolling_max方法计算最近N1个交易日的最高价, 最近N1个交易日的最高点: MaxIn_N1 对于上市不足N1天的数据，取上市至今的最高价 ''' index_data['MaxIn_N1'] = pandas.rolling_max(index_data['High'], N1) index_data['MaxIn_N1'].fillna(value=pandas.expanding_max(index_data['High']), inplace=True) ''' 通过rolling_min方法计算最近N2个交易日的最低价, MinIn_N2:MinIn_N2 ''' index_data['MinIn_N2'] = pandas.rolling_min(index_data['Low'], N1) index_data['MinIn_N2'].fillna(value=pandas.expanding_min(index_data['Low']), inplace=True) # 当当天的【close】> 昨天的【MaxIn_N1】时，将【收盘发出的信号】设定为1 buy_index = index_data[index_data['Close'] > index_data['MaxIn_N1'].shift(1)].index index_data.loc[buy_index, '收盘发出的信号'] = 1 # 当当天的【close】< 昨天的【MinIn_N2】时，将【收盘发出的信号】设定为0 sell_index = index_data[index_data['Close'] < index_data['MinIn_N2'].shift(1)].index index_data.loc[sell_index, '收盘发出的信号'] = 0 ''' The following sources are used for the test of the above function # 计算每天的仓位，当天持有上证指数时，仓位为1，当天不持有上证指数时，仓位为0 index_data['当天的仓位'] = index_data['收盘发出的信号'].shift(1) index_data['当天的仓位'].fillna(method='ffill', inplace=True)
data['fast_line']=pd.rolling_mean(data['open'],h)
data['slow_line']=pd.rolling_mean(data['open'],k)
data['fast_line']=data['fast_line'].fillna(value=pd.expanding_mean(data['open']))
data['slow_line']=data['slow_line'].fillna(value=pd.expanding_mean(data['open']))
data['dist_%s_%s'%(k,h)]=data['fast_line']-data['slow_line']
for i in range(5,31,5):
data['MA_%s'%i]=pd.rolling_mean(data['open'],i)
data['MA_%s'%i]=data['MA_%s'%i].fillna(0)-data['open']
data=MACD(data,12,26,9)
for h in range(10,26,5):
data['fast_line']=''
data['slow_line']=''
data['fast_line']=pd.rolling_max(data['high'].shift(1),h)
data['slow_line']=pd.rolling_min(data['low'].shift(1),h)
data['fast_line']=data['fast_line'].fillna(value=pd.expanding_max(data['high']))
data['slow_line']=data['slow_line'].fillna(value=pd.expanding_min(data['low']))
data['dist_high_%s'%h]=data['high']-data['fast_line']
data['dist_low_%s'%h]=data['low']-data['slow_line']
#引入隐马尔科夫模型
factor_list=['close','volume','dist_10_5','dist_15_5','dist_20_5','dist_15_10','dist_20_10','dist_20_15','dist_30_15','log_return','log_return_5','MACD','dist_high_10','dist_high_15','dist_high_20','dist_high_25','dist_low_10','dist_low_15','dist_low_20','dist_low_25','MA_5','MA_10','MA_15','MA_20','MA_25','MA_30']
for i in factor_list:
X = np.column_stack([data[i]])
model = GaussianHMM(n_components=3, covariance_type="diag", n_iter=1000,random_state=0).fit(X)
hidden_states = model.predict(X)
plt.figure(figsize=(15, 8))
for k in range(model.n_components):
idx = (hidden_states==k)
plt.plot_date(data['time'][idx],data['close'][idx],'.',label='%dth hidden state'%k,lw=1)
plt.legend()
plt.grid(1)
plt.savefig('C:/Users/Public/Documents/Python Scripts/隐马尔科夫状态刻画图集开盘价（2015）/%s.png'%(i))
for root, dirs, files in os.walk('all_trading_data/stock data'):
if files:
for f in files:
if '.csv' in f:
stock_code_list.append(f.split('.csv')[0])
all_stock = pd.DataFrame()
for code in stock_code_list:
print(code)
stock_data = pd.read_csv('all_trading_data/stock data/' + code + '.csv',
parse_dates=[1])
stock_data.sort('date', inplace=True)
low_list = pd.rolling_min(stock_data['low'], 9)
low_list.fillna(value=pd.expanding_min(stock_data['low']), inplace=True)
high_list = pd.rolling_max(stock_data['high'], 9)
high_list.fillna(value=pd.expanding_max(stock_data['high']), inplace=True)
rsv = (stock_data['close'] - low_list) / (high_list - low_list) * 100
stock_data['KDJ_K'] = pd.ewma(rsv, com=2)
stock_data['KDJ_D'] = pd.ewma(stock_data['KDJ_K'], com=2)
stock_data['KDJ_J'] = 3 * stock_data['KDJ_K'] - 2 * stock_data['KDJ_D']
stock_data['KDJ_金叉死叉'] = ''
kdj_position = stock_data['KDJ_K'] > stock_data['KDJ_D']
stock_data.loc[kdj_position[(kdj_position == True)
& (kdj_position.shitf() == False)].index,
'KDJ_金叉死叉'] = '金叉'
stock_data.loc[kdj_position[(kdj_position == False)
& (kdj_position.shitf() == True)].index,
'KDJ_金叉死叉'] = '死叉'
def LLV(self, param):
if param[1] == 0:
return pd.expanding_min(param[0])
return pd.rolling_min(param[0], param[1])
def longsklearn(code='999999', ptype='f',dtype='d',start=None,end=None):
# code='999999'
# dtype = 'w'
# start = '2014-09-01'
# start = None
# end='2015-12-23'
# end = None
df = tdd.get_tdx_append_now_df(code, ptype, start, end).sort_index(ascending=True)
# if not dtype == 'd':
# df = tdd.get_tdx_stock_period_to_type(df, dtype).sort_index(ascending=True)
dw = tdd.get_tdx_stock_period_to_type(df, dtype).sort_index(ascending=True)
# print df[:1]
h = df.loc[:, ['open', 'close', 'high', 'low']]
highp = h['high'].values
lowp = h['low'].values
openp = h['open'].values
closep = h['close'].values
lr = LinearRegression()
x = np.atleast_2d(np.linspace(0, len(closep), len(closep))).T
lr.fit(x, closep)
LinearRegression(copy_X=True, fit_intercept=True, n_jobs=1, normalize=False)
xt = np.atleast_2d(np.linspace(0, len(closep) + 200, len(closep) + 200)).T
yt = lr.predict(xt)
# plt.plot(xt,yt,'-g',linewidth=5)
# plt.plot(closep)
bV = []
bP = []
uV = []
uP = []
for i in range(1, len(highp) - 1):
# if highp[i] <= highp[i - 1] and highp[i] < highp[i + 1] and lowp[i] <= lowp[i - 1] and lowp[i] < lowp[i + 1]:
if lowp[i] <= lowp[i - 1] and lowp[i] < lowp[i + 1]:
bV.append(lowp[i])
bP.append(i)
for i in range(1, len(highp) - 1):
# if highp[i] >= highp[i - 1] and highp[i] > highp[i + 1] and lowp[i] >= lowp[i - 1] and lowp[i] > lowp[i + 1]:
if highp[i] >= highp[i - 1] and highp[i] > highp[i + 1]:
uV.append(highp[i])
uP.append(i)
print highp
print "uV:%s" % uV[:1]
print "uP:%s" % uP[:1]
print "bV:%s" % bV[:1]
print "bP:%s" % bP[:1]
sV, sP = LIS(uV)
dV, dP = LIS(bV)
print "sV:%s" % sV[:1]
print "sP:%s" % sP[:1]
print "dV:%s" % dV[:1]
print "dP:%s" % dP[:1]
sidx = []
didx = []
for i in range(len(sP)):
# idx.append(bP[p[i]])
sidx.append(uP[sP[i]])
for i in range(len(dP)):
# idx.append(bP[p[i]])
didx.append(bP[dP[i]])
print "sidx:%s"%sidx[:1]
print "didx:%s"%didx[:1]
# plt.plot(closep)
# plt.plot(idx,d,'ko')
lr = LinearRegression()
X = np.atleast_2d(np.array(sidx)).T
Y = np.array(sV)
lr.fit(X, Y)
estV = lr.predict(xt)
fig = plt.figure(figsize=(16, 10), dpi=72)
# plt.subplots_adjust(bottom=0.1, right=0.8, top=0.9)
plt.subplots_adjust(left=0.05, bottom=0.08, right=0.95, top=0.95, wspace=0.15, hspace=0.25)
# set (gca,'Position',[0,0,512,512])
# fig.set_size_inches(18.5, 10.5)
# fig=plt.fig(figsize=(14,8))
ax = fig.add_subplot(111)
plt.grid(True)
# print h.index[:5], h['close']
ax = h['close'].plot()
# ax.plot(pd.datetime(h.index),h['close'], linewidth=1)
# ax.plot(uP, uV, linewidth=1)
# ax.plot(uP, uV, 'ko')
# ax.plot(bP, bV, linewidth=1)
# ax.plot(bP, bV, 'bo')
# # ax.plot(sP, sV, linewidth=1)
# # ax.plot(sP, sV, 'yo')
# ax.plot(sidx, sV, linewidth=1)
# ax.plot(sidx, sV, 'ro')
# ax.plot(didx, dV, linewidth=1)
# ax.plot(didx, dV, 'co')
df['mean']=map(lambda h,l:(h+l)/2,df.high.values,df.low.values)
print df['mean'][:1]
# d=df.mean
dw=dw.set_index('date')
# print dw[:2]
# ax.plot(df.index,df['mean'],'g',linewidth=1)
ax.plot(df.index,pd.rolling_mean(df['mean'], 60), 'g',linewidth=1)
ax.plot(dw.index,pd.rolling_mean(dw.close, 5), 'r',linewidth=1)
ax.plot(dw.index,pd.rolling_min(dw.close, 5), 'bo')
ax.plot(dw.index,pd.rolling_max(dw.close, 5), 'yo')
ax.plot(dw.index,pd.expanding_max(dw.close, 5), 'ro')
ax.plot(dw.index,pd.expanding_min(dw.close, 5), 'go')
# print pd.rolling_min(df.close,20)[:1],pd.rolling_min(df.close,20)[-1:]
# print pd.rolling_min(df.close,20)
# print pd.rolling_max(df.close,20)[:1],pd.rolling_max(df.close,20)[-1:]
# print pd.rolling_max(df.close,20)
# ax.plot(idx, d, 'ko')
# ax.plot(xt, estV, '-r', linewidth=5)
# ax.plot(xt, yt, '-g', linewidth=5)
# ax2 = fig.add_subplot(122)
# print len(closep),len(idx),len(d),len(xt),len(estV),len(yt)
# f=lambda x:x[-int(len(x)/10):]
# ax2.plot(f(closep))
# ax2.plot(f(idx),f(d),'ko')
# ax2.plot(f(xt),f(estV),'-r',linewidth=5)
# ax2.plot(f(xt),f(yt),'-g',linewidth=5)
# # plt.show()
scale = 1.1
zp = zoompan.ZoomPan()
figZoom = zp.zoom_factory(ax, base_scale=scale)
figPan = zp.pan_factory(ax)
show()
# ==========计算海龟交易法则的买卖点
# 设定海龟交易法则的两个参数，当收盘价大于最近N1天的最高价时买入，当收盘价低于最近N2天的最低价时卖出
# 这两个参数可以自行调整大小，但是一般N1 > N2
N1 = 20
N2 = 10
# 通过rolling_max方法计算最近N1个交易日的最高价
index_data['最近N1个交易日的最高点'] = pd.rolling_max(index_data['high'], N1)
# 对于上市不足N1天的数据，取上市至今的最高价
index_data['最近N1个交易日的最高点'].fillna(value=pd.expanding_max(index_data['high']), inplace=True)
# 通过相似的方法计算最近N2个交易日的最低价
index_data['最近N2个交易日的最低点'] = pd.rolling_min(index_data['low'], N1)
index_data['最近N2个交易日的最低点'].fillna(value=pd.expanding_min(index_data['low']), inplace=True)
# 当当天的【close】> 昨天的【最近N1个交易日的最高点】时，将【收盘发出的信号】设定为1
buy_index = index_data[index_data['close'] > index_data['最近N1个交易日的最高点'].shift(1)].index
index_data.loc[buy_index, '收盘发出的信号'] = 1
# 当当天的【close】< 昨天的【最近N2个交易日的最低点】时，将【收盘发出的信号】设定为0
sell_index = index_data[index_data['close'] < index_data['最近N2个交易日的最低点'].shift(1)].index
index_data.loc[sell_index, '收盘发出的信号'] = 0
# 计算每天的仓位，当天持有上证指数时，仓位为1，当天不持有上证指数时，仓位为0
index_data['当天的仓位'] = index_data['收盘发出的信号'].shift(1)
index_data['当天的仓位'].fillna(method='ffill', inplace=True)
# 取1992年之后的数据，排出较早的数据
index_data = index_data[index_data['date'] >= pd.to_datetime('19930101')]
def comput_idicators(df,
trading_days,
required,
save_file,
save_address,
whole=1):
# TODO:net_value has some problem.
# columns needed
col = ['index_price', 'Interest_rate', 'nav', 'rebalancing', 'stoploss']
df_valid = df.ix[:, col]
start_balance = df.index[df['rebalancing'] == 1][0]
df_valid = df_valid[df_valid.index >= start_balance]
# daily return
df_valid['return'] = np.log(df['nav']) - np.log(df['nav'].shift(1))
# benchmark_net_value
df_valid[
'benchmark'] = df_valid['index_price'] / df_valid['index_price'].ix[0]
# benchmark_return
df_valid['benchmark_return'] = (df_valid['benchmark']-
df_valid['benchmark'].shift(1))/\
df_valid['benchmark'].shift(1)
# Annualized return
df_valid['Annu_return'] = pd.expanding_mean(
df_valid['return']) * trading_days
# Volatility
df_valid.loc[:, 'algo_volatility'] = pd.expanding_std(
df_valid['return']) * np.sqrt(trading_days)
df_valid.loc[:, 'xret'] = df_valid[
'return'] - df_valid['Interest_rate'] / trading_days / 100
df_valid.loc[:, 'ex_return'] = df_valid['return'] - df_valid[
'benchmark_return']
def ratio(x):
return np.nanmean(x) / np.nanstd(x)
# sharpe ratio
df_valid.loc[:, 'sharpe'] = pd.expanding_apply(df_valid['xret'], ratio)\
* np.sqrt(trading_days)
# information ratio
df_valid.loc[:, 'IR'] = pd.expanding_apply(df_valid['ex_return'], ratio)\
* np.sqrt(trading_days)
# Sortino ratio
def modify_ratio(x, re):
re /= trading_days
ret = np.nanmean(x) - re
st_d = np.nansum(np.square(x[x < re] - re)) / x[x < re].size
return ret / np.sqrt(st_d)
df_valid.loc[:, 'sortino'] = pd.expanding_apply(
df_valid['return'], modify_ratio,
args=(required, )) * np.sqrt(trading_days)
# Transfer infs to NA
df_valid.loc[np.isinf(df_valid.loc[:, 'sharpe']), 'sharpe'] = np.nan
df_valid.loc[np.isinf(df_valid.loc[:, 'IR']), 'IR'] = np.nan
# hit_rate
wins = np.where(df_valid['return'] >= df_valid['benchmark_return'], 1.0,
0.0)
df_valid.loc[:, 'hit_rate'] = wins.cumsum() / pd.expanding_apply(wins, len)
# 95% VaR
df_valid['VaR'] = -pd.expanding_quantile(df_valid['return'], 0.05)*\
np.sqrt(trading_days)
# 95% CVaR
df_valid['CVaR'] = -pd.expanding_apply(df_valid['return'],
lambda x: x[x < np.nanpercentile(x, 5)].mean())\
* np.sqrt(trading_days)
if whole == 1:
# max_drawdown
def exp_diff(x, type):
if type == 'dollar':
xret = pd.expanding_apply(x, lambda xx: (xx[-1] - xx.max()))
else:
xret = pd.expanding_apply(
x, lambda xx: (xx[-1] - xx.max()) / xx.max())
return xret
# dollar
# xret = exp_diff(df_valid['cum_profit'],'dollar')
# df_valid['max_drawdown_profit'] = abs(pd.expanding_min(xret))
# percentage
xret = exp_diff(df_valid['nav'], 'percentage')
df_valid['max_drawdown_ret'] = abs(pd.expanding_min(xret))
# max_drawdown_duration:
# drawdown_enddate is the first time for restoring the max
def drawdown_end(x, type):
xret = exp_diff(x, type)
minloc = xret[xret == xret.min()].index[0]
x_sub = xret[xret.index > minloc]
# if never recovering,then return nan
try:
return x_sub[x_sub == 0].index[0]
except:
return np.nan
def drawdown_start(x, type):
xret = exp_diff(x, type)
minloc = xret[xret == xret.min()].index[0]
x_sub = xret[xret.index < minloc]
try:
return x_sub[x_sub == 0].index[-1]
except:
return np.nan
df_valid['max_drawdown_start'] = pd.Series()
df_valid['max_drawdown_end'] = pd.Series()
df_valid['max_drawdown_start'].ix[-1] = drawdown_start(
df_valid['nav'], 'percentage')
df_valid['max_drawdown_end'].ix[-1] = drawdown_end(
df_valid['nav'], 'percentage')
df_valid.to_csv(save_address)
# =====result visualization=====
plt.figure(1)
if whole == 1:
plt.subplot(224)
plt.plot(df_valid['nav'], label='strategy')
plt.plot(df_valid['benchmark'], label='S&P500')
plt.xlabel('Date')
plt.legend(loc=0, shadow=True)
plt.ylabel('Nav')
plt.title('Nav of ' + save_file + ' & SP500')
# plt.subplot(223)
# plt.plot(df_valid['cum_profit'],label = 'strategy')
# plt.xlabel('Date')
# plt.ylabel('Cum_profit')
# plt.title('Cum_profit of ' + save_file)
plt.subplot(221)
plt.plot(df_valid['return'], label='strategy')
plt.xlabel('Date')
plt.ylabel('Daily_return')
plt.title('Daily Return of ' + save_file)
plt.subplot(222)
x_return = df_valid[df_valid['return'].notna()].loc[:, 'return']
y_return = df_valid[
df_valid['benchmark_return'].notna()].loc[:, 'benchmark_return']
mu = x_return.mean()
sigma = x_return.std()
mybins = np.linspace(mu - 3 * sigma, mu + 3 * sigma, 100)
count_x, _, _ = plt.hist(x_return,
mybins,
normed=1,
alpha=0.5,
label='strategy')
count_y, _, _ = plt.hist(y_return,
mybins,
normed=1,
alpha=0.5,
label='S&P500')
plt.ylabel('density')
plt.xlabel('daily_return')
plt.title('Histogram of Daily Return for ' + save_file + ' & SP500')
plt.grid(True)
# add normal distribution line
y = mlab.normpdf(mybins, mu, sigma)
plt.plot(mybins, y, 'r--', linewidth=1, label='Normal of strategy')
plt.legend(loc=0, shadow=True)
# plt.tight_layout()
plt.show()
return df_valid
df['n1_high'] = df['high'].rolling(window=N1,center=False).max() # print(df.head(43)) # 用pd.expanding_max()从第一个开始依次寻找目前出现过的最大值 # 实例如下 # demo_list = np.array([1,2,1,1,500,100]) # pd.expanding_max(demo_list) # array([ 1., 2., 2., 2., 500., 500.]) expan_max = pd.expanding_max(df['close']) #fillna() 将NaN替换为当前的序列 df['n1_high'].fillna(value=expan_max, inplace=True) # print(df.head(43)) df['n2_low'] = df['low'].rolling(window=N2,center=False).min() expan_min = pd.expanding_min(df['close']) df['n2_low'].fillna(value=expan_min, inplace=True) # print(df.head(22)) ##做空的序列 df['n1_low'] = df['low'].rolling(window=N1,center=False).min() df['n1_low'].fillna(value=expan_min, inplace=True) df['n2_high'] = df['high'].rolling(window=N2,center=False).max() df['n2_high'].fillna(value=expan_max, inplace=True) #接下来根据突破的定义来构建signal列 #当天收盘价格超过N天最高或最低价，超过最高价是作为买入信号 #buy_index=行的索引，本例中就是日期，而且是满足close大于n1_high的情况下的索引 buy_index_in = df[df['close'] > df['n1_high'].shift(1)].index #shift 是移动序列，即将整个n1_high列都下移一格
# 2014-05-28 210.24 212.77 205.26 -0.624 210.02 211.56 5496278 # 2014-05-29 210.24 212.49 207.72 0.000 210.57 210.24 3694596 # 2014-05-30 207.77 214.80 207.02 -1.175 210.30 210.24 5586068 # 2014-06-02 204.70 209.35 201.67 -1.478 207.33 207.77 4668115 # 2014-06-03 204.94 208.00 202.59 0.117 203.49 204.70 3866182 # # date date_week atr21 atr14 key n1_high # 2014-05-28 20140528 2 7.5100 7.5100 0 210.24 # 2014-05-29 20140529 3 6.0748 6.0421 1 210.24 # 2014-05-30 20140530 4 6.6981 6.7060 2 210.24 # 2014-06-02 20140602 0 7.2350 7.2763 3 210.24 # 2014-06-03 20140603 1 6.7973 6.7894 4 210.24 #下面使用类似的方式构建N2天内最低价格卖出信号n2_low: #rolling_min()函数和rolling_max()函数类似 tsla_df['n2_low'] = pd.rolling_min(tsla_df['low'], window=N2) expan_min = pd.expanding_min(tsla_df['close']) tsla_df['n2_low'].fillna(value=expan_min, inplace=True) #下面根据突破的定义来构建signal列： #当天的收盘价格超过N天内的最高价或最低价，超过最高价格作为买入信号买入股票持有 buy_index = tsla_df[tsla_df['close'] > tsla_df['n1_high'].shift(1)].index tsla_df.loc[buy_index, 'signal'] = 1 #当天收盘价格超过N天内的最高价格或最低价格，超过最低价格作为卖出信号 sell_index = tsla_df[tsla_df['close'] < tsla_df['n2_low'].shift(1)].index tsla_df.loc[sell_index, 'signal'] = 0 #筛选条件 今天的收盘价格>截止到昨天的最高价格 和 今天的收盘价格 < 截止到昨天的最低价格 #下面使用饼图显示在整个交易中信号的产生情况，可以发现买入信号比卖出信号多 #如下图所示 tsla_df.signal.value_counts().plot(kind='pie', figsize=(5, 5)) # 1.0 54 # 0.0 53 # Name: signal, dtype: int64
def analyse(self):
# Logger.log(logging.INFO, "Analyse Strategy", {"scope":__name__, "Rule 1":self._rule1, "Rule 2":self._rule2, "Rule 3":self._rule3, "Type":self._type})
connection = sqlite3.connect(pyswing.database.pySwingDatabase)
query = self.analyseStrategySql % (self._rule1, self._rule2,
self._rule3, self._exit, self._type)
self._strategyData = read_sql_query(query, connection, 'Date')
self._strategyData[
'ExitValueAfterCosts'] = self._strategyData['ExitValue'] - 0.2
connection.close()
exitValueDataFrame = self._strategyData.ix[:, 'ExitValueAfterCosts']
mean = exitValueDataFrame.mean()
median = exitValueDataFrame.median()
sum = exitValueDataFrame.sum()
count = exitValueDataFrame.count()
tradesPerYear = count / 10
sharpeRatio = sqrt(tradesPerYear) * exitValueDataFrame.mean(
) / exitValueDataFrame.std()
self._strategyData["Sum"] = expanding_sum(exitValueDataFrame)
self._strategyData["Max"] = expanding_max(self._strategyData["Sum"])
self._strategyData["Min"] = expanding_min(self._strategyData["Sum"])
self._strategyData[
"DD"] = self._strategyData["Max"] - self._strategyData["Min"]
runningSum = expanding_sum(exitValueDataFrame)
max2here = expanding_max(runningSum)
dd2here = runningSum - max2here
drawDown = dd2here.min()
Logger.log(
logging.INFO, "Analysing Strategy", {
"scope": __name__,
"Rule 1": self._rule1,
"Rule 2": self._rule2,
"Rule 3": self._rule3,
"Exit": self._exit,
"Type": self._type,
"Mean": str(mean),
"Median": str(median),
"Sum": str(sum),
"Count": str(count),
"SharpeRatio": str(sharpeRatio),
"DrawDown": str(drawDown)
})
connection = sqlite3.connect(pyswing.database.pySwingDatabase)
c = connection.cursor()
deleteSql = self.deleteStrategySql % (
pyswing.globals.pySwingStrategy, self._rule1, self._rule2,
self._rule3, self._exit, self._type)
c.executescript(deleteSql)
connection.commit()
insertSql = self.insertStrategySql % (
pyswing.globals.pySwingStrategy, self._rule1, self._rule2,
self._rule3, self._exit, self._type, str(mean), str(median),
str(sum), str(count), str(sharpeRatio), str(drawDown))
c.executescript(insertSql)
connection.commit()
c.close()
connection.close()
if __name__ == '__main__':
kl_pd = ABuSymbolPd.make_kl_df('TSLA', n_folds=2)
# 1、这里采用N日趋势突破，即超过N1天内的最高价，就买入，低于N2天内的最低价，就卖出
N1 = 42
N2 = 21
# 2.1 采用pd.rolling_max可以寻找一个窗口长度内最大值
kl_pd['n1_high'] = pd.rolling_max(kl_pd['high'], window=N1)
# 2.2 但这样会导致前N1-1个元素为NAN，
# 我们使用pd.expanding_max来填充NAN，expanding_max会逐个遍历数组元素，并把返回直到当前位置看到过的最大元素
# 用前k天的收盘价来代替，k∈[0,N1]
expan_max = pd.expanding_max(kl_pd['close'])
kl_pd['n1_high'].fillna(expan_max, inplace=True)
# 2.3 最小值同理
kl_pd['n2_low'] = pd.rolling_min(kl_pd['low'], window=N2)
expan_min = pd.expanding_min(kl_pd['close'])
kl_pd['n2_low'].fillna(expan_min, inplace=True)
print(kl_pd.head())
# 3.1 根据n1_high和n2_low来定义买入卖出的信号序列
# 注意，是当天的收盘价，高于昨天以前的n1值，就买入，不能包括今天的，因为今天的收盘价，怎么也不会高于今天的最高值
buy_signal = kl_pd[kl_pd.close > kl_pd.n1_high.shift(1)].index
kl_pd.loc[buy_signal, 'signal'] = 1
# 3.2 n2_low的卖出信号同理
sell_signal = kl_pd[kl_pd.close < kl_pd.n2_low.shift(1)].index
kl_pd.loc[sell_signal, 'signal'] = 0
# 3.3 这里可以不用考虑Nan的情况
kl_pd.signal.value_counts().plot(kind='pie')
plt.show()
# 4.1 将买入卖出的信号转化为持股的状态
def longsklearn(code='999999', ptype='f', dtype='d', start=None, end=None):
# code='999999'
# dtype = 'w'
# start = '2014-09-01'
# start = None
# end='2015-12-23'
# end = None
df = tdd.get_tdx_append_now_df(code, ptype, start,
end).sort_index(ascending=True)
# if not dtype == 'd':
# df = tdd.get_tdx_stock_period_to_type(df, dtype).sort_index(ascending=True)
dw = tdd.get_tdx_stock_period_to_type(df, dtype).sort_index(ascending=True)
# print df[:1]
h = df.loc[:, ['open', 'close', 'high', 'low']]
highp = h['high'].values
lowp = h['low'].values
openp = h['open'].values
closep = h['close'].values
lr = LinearRegression()
x = np.atleast_2d(np.linspace(0, len(closep), len(closep))).T
lr.fit(x, closep)
LinearRegression(copy_X=True,
fit_intercept=True,
n_jobs=1,
normalize=False)
xt = np.atleast_2d(np.linspace(0, len(closep) + 200, len(closep) + 200)).T
yt = lr.predict(xt)
# plt.plot(xt,yt,'-g',linewidth=5)
# plt.plot(closep)
bV = []
bP = []
uV = []
uP = []
for i in range(1, len(highp) - 1):
# if highp[i] <= highp[i - 1] and highp[i] < highp[i + 1] and lowp[i] <= lowp[i - 1] and lowp[i] < lowp[i + 1]:
if lowp[i] <= lowp[i - 1] and lowp[i] < lowp[i + 1]:
bV.append(lowp[i])
bP.append(i)
for i in range(1, len(highp) - 1):
# if highp[i] >= highp[i - 1] and highp[i] > highp[i + 1] and lowp[i] >= lowp[i - 1] and lowp[i] > lowp[i + 1]:
if highp[i] >= highp[i - 1] and highp[i] > highp[i + 1]:
uV.append(highp[i])
uP.append(i)
print(highp)
print("uV:%s" % uV[:1])
print("uP:%s" % uP[:1])
print("bV:%s" % bV[:1])
print("bP:%s" % bP[:1])
sV, sP = LIS(uV)
dV, dP = LIS(bV)
print("sV:%s" % sV[:1])
print("sP:%s" % sP[:1])
print("dV:%s" % dV[:1])
print("dP:%s" % dP[:1])
sidx = []
didx = []
for i in range(len(sP)):
# idx.append(bP[p[i]])
sidx.append(uP[sP[i]])
for i in range(len(dP)):
# idx.append(bP[p[i]])
didx.append(bP[dP[i]])
print("sidx:%s" % sidx[:1])
print("didx:%s" % didx[:1])
# plt.plot(closep)
# plt.plot(idx,d,'ko')
lr = LinearRegression()
X = np.atleast_2d(np.array(sidx)).T
Y = np.array(sV)
lr.fit(X, Y)
estV = lr.predict(xt)
fig = plt.figure(figsize=(16, 10), dpi=72)
# plt.subplots_adjust(bottom=0.1, right=0.8, top=0.9)
plt.subplots_adjust(left=0.05,
bottom=0.08,
right=0.95,
top=0.95,
wspace=0.15,
hspace=0.25)
# set (gca,'Position',[0,0,512,512])
# fig.set_size_inches(18.5, 10.5)
# fig=plt.fig(figsize=(14,8))
ax = fig.add_subplot(111)
plt.grid(True)
# print h.index[:5], h['close']
ax = h['close'].plot()
# ax.plot(pd.datetime(h.index),h['close'], linewidth=1)
# ax.plot(uP, uV, linewidth=1)
# ax.plot(uP, uV, 'ko')
# ax.plot(bP, bV, linewidth=1)
# ax.plot(bP, bV, 'bo')
# # ax.plot(sP, sV, linewidth=1)
# # ax.plot(sP, sV, 'yo')
# ax.plot(sidx, sV, linewidth=1)
# ax.plot(sidx, sV, 'ro')
# ax.plot(didx, dV, linewidth=1)
# ax.plot(didx, dV, 'co')
df['mean'] = list(
map(lambda h, l: (h + l) / 2, df.high.values, df.low.values))
print(df['mean'][:1])
# d=df.mean
dw = dw.set_index('date')
# print dw[:2]
# ax.plot(df.index,df['mean'],'g',linewidth=1)
ax.plot(df.index, pd.rolling_mean(df['mean'], 60), 'g', linewidth=1)
ax.plot(dw.index, pd.rolling_mean(dw.close, 5), 'r', linewidth=1)
ax.plot(dw.index, pd.rolling_min(dw.close, 5), 'bo')
ax.plot(dw.index, pd.rolling_max(dw.close, 5), 'yo')
ax.plot(dw.index, pd.expanding_max(dw.close, 5), 'ro')
ax.plot(dw.index, pd.expanding_min(dw.close, 5), 'go')
# print pd.rolling_min(df.close,20)[:1],pd.rolling_min(df.close,20)[-1:]
# print pd.rolling_min(df.close,20)
# print pd.rolling_max(df.close,20)[:1],pd.rolling_max(df.close,20)[-1:]
# print pd.rolling_max(df.close,20)
# ax.plot(idx, d, 'ko')
# ax.plot(xt, estV, '-r', linewidth=5)
# ax.plot(xt, yt, '-g', linewidth=5)
# ax2 = fig.add_subplot(122)
# print len(closep),len(idx),len(d),len(xt),len(estV),len(yt)
# f=lambda x:x[-int(len(x)/10):]
# ax2.plot(f(closep))
# ax2.plot(f(idx),f(d),'ko')
# ax2.plot(f(xt),f(estV),'-r',linewidth=5)
# ax2.plot(f(xt),f(yt),'-g',linewidth=5)
# # plt.show()
scale = 1.1
zp = zoompan.ZoomPan()
figZoom = zp.zoom_factory(ax, base_scale=scale)
figPan = zp.pan_factory(ax)
show()
# ========== 根据上一步得到的代码列表，遍历所有股票，将这些股票合并到一张表格all_stock中
all_stock = pd.DataFrame()
# 遍历每个创业板的股票
for code in stock_code_list:
print code
# 从csv文件中读取该股票数据
stock_data = pd.read_csv('trading-data@full/stock data/' + code + '.csv',
parse_dates=[1])# 注意：这里请填写数据文件在您电脑中的路径
stock_data.sort('date', inplace=True)# 对数据按照【date】交易日期从小到大排序
# 计算KDJ指标
low_list = pd.rolling_min(stock_data['low'], 9)
low_list.fillna(value=pd.expanding_min(stock_data['low']), inplace=True)
high_list = pd.rolling_max(stock_data['high'], 9)
high_list.fillna(value=pd.expanding_max(stock_data['high']), inplace=True)
rsv = (stock_data['close'] - low_list) / (high_list - low_list) * 100
stock_data['KDJ_K'] = pd.ewma(rsv, com=2)
stock_data['KDJ_D'] = pd.ewma(stock_data['KDJ_K'], com=2)
stock_data['KDJ_J'] = 3 * stock_data['KDJ_K'] - 2 * stock_data['KDJ_D']
# 计算KDJ指标金叉、死叉情况
stock_data['KDJ_金叉死叉'] = ''
kdj_position = stock_data['KDJ_K'] > stock_data['KDJ_D']
stock_data.loc[kdj_position[(kdj_position == True) & (kdj_position.shift() == False)].index, 'KDJ_金叉死叉'] = '金叉'
stock_data.loc[kdj_position[(kdj_position == False) & (kdj_position.shift() == True)].index, 'KDJ_金叉死叉'] = '死叉'
# 通过复权价格计算接下来几个交易日的收益率
for n in [1, 2, 3, 5, 10, 20]:
stock_data['接下来'+str(n)+'个交易日涨跌幅'] = stock_data['adjust_price'].shift(-1*n) / stock_data['adjust_price'] - 1.0
