def Chaikin_vol(High, Low, timeperiod=10):
Range = High - Low
return ((pd.DataFrame(talib.EMA(Range, timeperiod)).pct_change() - 1) *
100.0).values
import pandas as pd
import pandas.io.data as web
import datetime as dt
import matplotlib.pyplot as plt
import talib as ta
# Data Download
start = dt.datetime(2014, 10, 1)
end = dt.datetime(2015, 9, 30)
aapl = web.DataReader('AAPL', 'yahoo', start, end)
##########
# 2. Exponential Moving Averages EMA(5 & 21) Calculation and Chart
# Technical Indicator Calculation
aapl['ema5'] = ta.EMA(np.asarray(aapl['Close']), 5)
aapl['ema21'] = ta.EMA(np.asarray(aapl['Close']), 21)
# Technical Indicator Chart
aapl.plot(y=['Close', 'ema5', 'ema21'])
plt.title('Apple Close Prices & Exponential Moving Averages EMA(5 & 21)')
plt.legend(loc='upper left')
plt.show()
##########
# 3. Double Crossover Trading Signals
# Previous Periods Data (avoid back-testing bias)
aapl['ema5(-1)'] = aapl['ema5'].shift(1)
aapl['ema21(-1)'] = aapl['ema21'].shift(1)
aapl['ema5(-2)'] = aapl['ema5'].shift(2)
aapl['ema21(-2)'] = aapl['ema21'].shift(2)
def __uptrend(price_action, window_size):
ema = ta.EMA(price_action, window_size)
for i in range(1, window_size + 1):
if price_action[-i] < ema[-i]:
return False
return True
# AVGPRICE
ti_AVGPRICE = talib.AVGPRICE(bars['open'].values, bars['high'].values,
bars['low'].values, bars['close'].values)
# CMO
ti_CMO = talib.CMO(bars['price'].values, time_period)
# DEMA
ti_DEMA = talib.DEMA(bars['price'].values, time_period)
# DX
ti_DX = talib.DX(bars['high'].values, bars['low'].values, bars['close'].values,
time_period)
# EMA
ti_EMA = talib.EMA(bars['price'].values, time_period)
# MACDEXT
ti_MACD, ti_MACDSIGNAL, ti_MACDHIST = talib.MACDFIX(bars['price'].values)
# MEDPRICE
ti_MEDPRICE = talib.MEDPRICE(bars['high'].values, bars['low'].values)
# MFI
ti_MFI = talib.MFI(bars['high'].values, bars['low'].values,\
bars['close'].values, vol['volume'].values, time_period)
# MINUS_DI
ti_MINUS_DI = talib.MINUS_DI(bars['high'].values, bars['low'].values,\
bars['close'].values, time_period)
i = 0
list_indicators = []
list_missing = []
for index, row in final_pairs_to_csv.iterrows():
try:
i += 1
df_daily = cryptocompare_wrapper.hourly_price_historical(row['Coin'],
row['Pair_tuple'], limit=250, aggregate=4, exchange=row['Exchange'])
closes = df_daily.close.values
lows = df_daily.low.values
highs = df_daily.high.values
opens = df_daily.open.values
volumetos = df_daily.volumeto.values
rsi = talib.RSI(closes, timeperiod=14)
ema12 = talib.EMA(closes, timeperiod=12)
ema26 = talib.EMA(closes, timeperiod=26)
ema50 = talib.EMA(closes, timeperiod=50)
ema200 = talib.EMA(closes, timeperiod=200)
atr = talib.ATR(highs, lows, closes, timeperiod=14)
obv = talib.OBV(closes, volumetos)
list_indicators.append([row.Coin + '/' + row.Pair_tuple, closes[-2], rsi[-2], ema50[-2], ema200[-2]])
print(i)
#if i == 5: break
except:
list_missing.append([row.Coin + '/' + row.Pair_tuple, row.Exchange])
pass
df_indicators = pd.DataFrame(list_indicators, columns=['Pair', '4H Close', '4H RSI', '4H EMA50', '4H EMA200'])
df_indicators.set_index('Pair', inplace=True)
def EMA(matrix, timeperiod):
matrix['EMA' + str(timeperiod)] = talib.EMA(real=matrix['1-Close'].values,
timeperiod=timeperiod)
return matrix
def main():
LogProfitReset()
LogReset()
Log("init OK", time.strftime('%Y-%m-%d %X', time.localtime(time.time())))
Log(a, b, c, d)
_G("ok", 123)
Log(GetPid(), _G(), _G("ok"), _G("dummy"))
Sleep(1000)
_G(None)
Log(_G("ok"))
LogStatus("Time", time.time())
EnableLog(True)
SetErrorFilter("net")
Log(GetLastError())
Log(GetCommand())
ticker = exchange.GetTicker()
Log('ticker buy', ticker.Buy, ticker['Buy'])
r = _C(exchange.GetRecords)
Log(TA.ATR(r))
Log(TA.EMA(r, 10))
# test talib
Log(str(talib.EMA(r.Close, 10)))
for e in exchanges:
Log(e.GetName(), e.GetRate(), e.GetCurrency(), e.GetFee())
Log(e.GetAccount())
Log(_C(e.GetOrders))
Log(e.GetOrder(10))
Log(e.CancelOrder(10000))
Log(e.IO("api", "userinfo"))
Log(e.IO("api", "order_info", "symbol=btc_usd&order_id=3"))
Log(e.GetUSDCNY())
#Log(e.GetPosition())
#Log(e.SetContractType("next_week"))
Log(e.GetTicker())
Log('Asks:', len(e.GetDepth().Asks))
#Log(e.SetMarginLevel(10))
#Log(e.SetDirection("buy"))
#Log(e.SetContractType("quarter"))
#Log(e.GetRecords(PERIOD_M30)[0])
Log(e.GetRecords()[0])
x = Chart({
'title': {
'text': 'test chart'
},
'xAxis': {
'type': 'datetime'
},
'series': [{
'name': 'Buy',
'data': []
}, {
'name': 'Sell',
'data': []
}]
})
x.reset()
Log("策略将每10秒更新一次ticker")
for i in range(100):
ts = int(time.time() * 1000)
ticker = _C(exchange.GetTicker)
x.add(0, [ts, ticker.Buy])
x.add(1, [ts, ticker.Sell])
LogStatus(ticker)
Sleep(10000)
def vfi(dataframe, length=130, coef=0.2, vcoef=2.5, signalLength=5, smoothVFI=False):
"""
Volume Flow Indicator conversion
Author: creslinux, June 2018 - Python
Original Author: Chris Moody, TradingView - Pinescript
To return vfi, vfima and histogram
A simplified interpretation of the VFI is:
* Values above zero indicate a bullish state and the crossing of the zero line is the trigger or buy signal.
* The strongest signal with all money flow indicators is of course divergence.
* A crossover of vfi > vfima is uptrend
* A crossunder of vfima > vfi is downtrend
* smoothVFI can be set to smooth for a cleaner plot to ease false signals
* histogram can be used against self -1 to check if upward or downward momentum
Call from strategy to populate vfi, vfima, vfi_hist into dataframe
Example how to call:
# Volume Flow Index: Add VFI, VFIMA, Histogram to DF
dataframe['vfi'], dataframe['vfima'], dataframe['vfi_hist'] = \
vfi(dataframe, length=130, coef=0.2, vcoef=2.5, signalLength=5, smoothVFI=False)
:param dataframe:
:param length: - VFI Length - 130 default
:param coef: - price coef - 0.2 default
:param vcoef: - volume coef - 2.5 default
:param signalLength: - 5 default
:param smoothVFI: bool - False detault
:return: vfi, vfima, vfi_hist
"""
""""
Original Pinescript
From: https://www.tradingview.com/script/MhlDpfdS-Volume-Flow-Indicator-LazyBear/
length = input(130, title="VFI length")
coef = input(0.2)
vcoef = input(2.5, title="Max. vol. cutoff")
signalLength=input(5)
smoothVFI=input(false, type=bool)
#### Conversion summary to python
- ma(x,y) => smoothVFI ? sma(x,y) : x // Added as smoothVFI test on vfi
- typical = hlc3 // Added to DF as HLC
- inter = log(typical) - log(typical[1]) // Added to DF as inter
- vinter = stdev(inter, 30) // Added to DF as vinter
- cutoff = coef * vinter * close // Added to DF as cutoff
- vave = sma(volume, length)[1] // Added to DF as vave
- vmax = vave * vcoef // Added to Df as vmax
- vc = iff(volume < vmax, volume, vmax) // Added np.where test, result in DF as vc
- mf = typical - typical[1] // Added into DF as mf - typical is hlc3
- vcp = iff(mf > cutoff, vc, iff(mf < -cutoff, -vc, 0)) // added in def vcp, in DF as vcp
- vfi = ma(sum(vcp, length) / vave, 3) // Added as DF vfi. Will sma vfi 3 if smoothVFI flag set
- vfima = ema(vfi, signalLength) // added to DF as vfima
- d = vfi-vfima // Added to df as histogram
### Pinscript plotout - nothing to do here for freqtrade.
plot(0, color=gray, style=3)
showHisto=input(false, type=bool)
plot(showHisto ? d : na, style=histogram, color=gray, linewidth=3, transp=50)
plot( vfima , title="EMA of vfi", color=orange)
plot( vfi, title="vfi", color=green,linewidth=2)
"""
import talib as ta
from math import log
from pyti.simple_moving_average import simple_moving_average as sma
from numpy import where
length = length
coef = coef
vcoef = vcoef
signalLength = signalLength
smoothVFI = smoothVFI
df = dataframe
# Add hlc3 and populate inter to the dataframe
df['hlc'] = ((df['high'] + df['low'] + df['close']) / 3).astype(float)
df['inter'] = df['hlc'].map(log) - df['hlc'].shift(+1).map(log)
df['vinter'] = df['inter'].rolling(30).std(ddof=0)
df['cutoff'] = (coef * df['vinter'] * df['close'])
# Vave is to be calculated on volume of the past bar
df['vave'] = sma(df['volume'].shift(+1), length)
df['vmax'] = df['vave'] * vcoef
df['vc'] = where((df['volume'] < df['vmax']), df['volume'], df['vmax'])
df['mf'] = df['hlc'] - df['hlc'].shift(+1)
# more logic for vcp, so create a def and df.apply it
def vcp(x):
if x['mf'] > x['cutoff']:
return x['vc']
elif x['mf'] < -(x['cutoff']):
return -(x['vc'])
else:
return 0
df['vcp'] = df.apply(vcp, axis=1)
# vfi has a smooth option passed over def call, sma if set
df['vfi'] = (df['vcp'].rolling(length).sum()) / df['vave']
if smoothVFI == True:
df['vfi'] = sma(df['vfi'], 3)
df['vfima'] = ta.EMA(df['vfi'], signalLength)
df['vfi_hist'] = df['vfi'] - df['vfima']
# clean up columns used vfi calculation but not needed for strat
df.drop('hlc', axis=1, inplace=True)
df.drop('inter', axis=1, inplace=True)
df.drop('vinter', axis=1, inplace=True)
df.drop('cutoff', axis=1, inplace=True)
df.drop('vave', axis=1, inplace=True)
df.drop('vmax', axis=1, inplace=True)
df.drop('vc', axis=1, inplace=True)
df.drop('mf', axis=1, inplace=True)
df.drop('vcp', axis=1, inplace=True)
return df['vfi'], df['vfima'], df['vfi_hist']
def on_tick(self, tick: TickData):
"""
Callback of new tick data update.
"""
# 时间过滤
if (tick.datetime.time() > self.day_open_time
and tick.datetime.time() < self.day_close_time) or (
tick.datetime.time() > self.night_open_time) or (
tick.datetime.time() < self.night_close_time):
self.flag = 1
self.put_counter += 1
if self.put_counter > 120:
self.put_counter = 0
self.put_event()
else:
if self.flag == 0:
return
tick_time = tick.datetime
self.write_log("{0} in on_tick:不在交易时间".format(tick_time))
print("{0} in on_tick:不在交易时间".format(tick_time))
# 活跃单全撤
self.cancel_all()
# 市价平仓
if self.get_pos(self.leg1_symbol) > 0:
# leg1平多
vt_orderids = self.sell(
self.leg1_symbol, self.leg1_last_bid - 10 * self.pricetick,
abs(self.get_pos(self.leg1_symbol)))
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
if self.get_pos(self.leg1_symbol) < 0:
# leg1平空
vt_orderids = self.cover(
self.leg1_symbol, self.leg1_last_ask + 10 * self.pricetick,
abs(self.get_pos(self.leg1_symbol)))
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
if self.get_pos(self.leg2_symbol) > 0:
# leg2平多
vt_orderids = self.sell(
self.leg2_symbol, self.leg2_last_bid - 10 * self.pricetick,
abs(self.get_pos(self.leg2_symbol)))
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
if self.get_pos(self.leg2_symbol) < 0:
# leg2平空
vt_orderids = self.cover(
self.leg2_symbol, self.leg2_last_ask + 10 * self.pricetick,
abs(self.get_pos(self.leg2_symbol)))
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
self.flag = 0
# 回到初始状态
self.write_log("该交易时段交易次数{0},单腿风险次数{1},单腿风险比{2}".format(
self.trade_counter, self.risk_counter,
2 * self.risk_counter / self.trade_counter + 1))
print("该交易时段交易次数{0},单腿风险次数{1},单腿风险比{2}".format(
self.trade_counter, self.risk_counter,
2 * self.risk_counter / self.trade_counter + 1))
self.spread_count = 0
self.spread_mid_array.clear()
self.counter = 0
self.try_open = 0
self.risk_counter = 0
self.trade_counter = 0
self.put_counter = 0
return
# 行情
if tick.vt_symbol == self.leg1_symbol:
self.leg1_last_ask = tick.ask_price_1
self.leg1_last_bid = tick.bid_price_1
if self.dom_symbol == 1:
spread_mid = (self.leg1_last_ask - self.leg2_last_bid +
self.leg1_last_bid - self.leg2_last_ask) / 2
self.spread_mid_array.append(spread_mid)
self.spread_count += 1
else:
return
elif tick.vt_symbol == self.leg2_symbol:
self.leg2_last_ask = tick.ask_price_1
self.leg2_last_bid = tick.bid_price_1
if self.dom_symbol == 2:
spread_mid = (self.leg1_last_ask - self.leg2_last_bid +
self.leg1_last_bid - self.leg2_last_ask) / 2
self.spread_mid_array.append(spread_mid)
self.spread_count += 1
else:
return
else:
return
# 确保行情足够
if self.spread_count < self.ema_window + 5:
return
# 检查委托
if self.active_orderids:
if self.try_open:
# 试图开仓中，判断是否存在一边全成
if (abs(self.get_pos(self.leg1_symbol))
== self.arb_size) or (abs(
self.get_pos(self.leg2_symbol)) == self.arb_size):
if abs(self.get_pos(self.leg1_symbol)) != self.arb_size:
tick_time = tick.datetime
code_time = datetime.now()
self.write_log(
"code{0} tick{1} in on_tick:{2}尝试追单".format(
code_time, tick_time, self.leg1_symbol))
#print("code{0} tick{1} in on_tick:{2}尝试追单".format(code_time, tick_time, self.leg1_symbol))
# leg1没到位
for vt_orderid in list(self.active_orderids):
order = self.get_order(vt_orderid)
if not order:
self.write_log("未找到该订单{}".format(vt_orderid))
if order and order.vt_symbol == self.leg1_symbol:
tar_direction = order.direction
tar_offset = order.offset
tar_price = order.price
tar_vol = self.arb_size - abs(
self.get_pos(self.leg1_symbol))
if tar_direction == Direction.LONG:
tar_price += self.price_add * self.pricetick
else:
tar_price -= self.price_add * self.pricetick
self.cancel_order(order.vt_orderid)
# 挂新单
vt_orderids = self.send_order(
order.vt_symbol, tar_direction, tar_offset,
tar_price, tar_vol)
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
tick_time = tick.datetime
code_time = datetime.now()
self.write_log(
"code{} tick{} in on_tick:追单完成".format(
code_time, tick_time))
#print("code{} tick{} in on_tick:追单完成".format(code_time,tick_time))
self.try_open = 0
break
if abs(self.get_pos(self.leg2_symbol)) != self.arb_size:
tick_time = tick.datetime
code_time = datetime.now()
self.write_log(
"code{} tick{} in on_tick:{}尝试追单".format(
code_time, tick_time, self.leg2_symbol))
#print("code{} tick{} in on_tick:{}尝试追单".format(code_time,tick_time,self.leg2_symbol))
# leg2没到位
for vt_orderid in list(self.active_orderids):
order = self.get_order(vt_orderid)
if not order:
self.write_log("未找到该订单{}".format(vt_orderid))
if order and order.vt_symbol == self.leg2_symbol:
tar_direction = order.direction
tar_offset = order.offset
tar_price = order.price
tar_vol = self.arb_size - abs(
self.get_pos(self.leg2_symbol))
if tar_direction == Direction.LONG:
tar_price += self.price_add * self.pricetick
else:
tar_price -= self.price_add * self.pricetick
self.cancel_order(order.vt_orderid)
# 挂新单
vt_orderids = self.send_order(
order.vt_symbol, tar_direction, tar_offset,
tar_price, tar_vol)
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
tick_time = tick.datetime
code_time = datetime.now()
self.write_log(
"code{} tick{} in on_tick:追单完成".format(
code_time, tick_time))
#print("code{} tick{} in on_tick:追单完成".format(code_time,tick_time))
self.try_open = 0
break
if self.counter < self.time_out:
self.counter += 1
return
tick_time = tick.datetime
code_time = datetime.now()
self.write_log(
"code{} tick{} in on_tick:等待时间到，当前仓位：近月{}远月{}，本地委托{}".format(
code_time, tick_time, self.get_pos(self.leg1_symbol),
self.get_pos(self.leg2_symbol), self.active_orderids))
#print("code{} tick{} in on_tick:等待时间到，当前仓位：近月{}远月{}，本地委托{}".format(code_time, tick_time,self.get_pos(self.leg1_symbol),self.get_pos(self.leg2_symbol),self.active_orderids))
self.try_open = 0
# 委托全撤
self.cancel_all()
return
else:
self.try_open = 0
self.counter = 0
# 检查单腿风险
if (self.get_pos(self.leg1_symbol) +
self.get_pos(self.leg2_symbol)) != 0:
tick_time = tick.datetime
code_time = datetime.now()
self.write_log(
"code{} tick{} in on_tick:出现单腿风险，当前仓位：近月{}远月{}，本地委托{}".format(
code_time, tick_time, self.get_pos(self.leg1_symbol),
self.get_pos(self.leg2_symbol), self.active_orderids))
# 记录启动以来总单腿风险次数
self.risk_counter += 1
# 处理单腿
# 市价平仓
if self.get_pos(self.leg1_symbol) > 0:
# leg1平多
vt_orderids = self.sell(
self.leg1_symbol, self.leg1_last_bid - 10 * self.pricetick,
abs(self.get_pos(self.leg1_symbol)))
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
if self.get_pos(self.leg1_symbol) < 0:
# leg1平空
vt_orderids = self.cover(
self.leg1_symbol, self.leg1_last_ask + 10 * self.pricetick,
abs(self.get_pos(self.leg1_symbol)))
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
if self.get_pos(self.leg2_symbol) > 0:
# leg2平多
vt_orderids = self.sell(
self.leg2_symbol, self.leg2_last_bid - 10 * self.pricetick,
abs(self.get_pos(self.leg2_symbol)))
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
if self.get_pos(self.leg2_symbol) < 0:
# leg2平空
vt_orderids = self.cover(
self.leg2_symbol, self.leg2_last_ask + 10 * self.pricetick,
abs(self.get_pos(self.leg2_symbol)))
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
return
# 区间更新
ema = talib.EMA(np.array(self.spread_mid_array),
timeperiod=self.ema_window)
self.ema_mid = ema[-2]
self.ema_up = ema[-2] + self.ema_th * self.pricetick
self.ema_down = ema[-2] - self.ema_th * self.pricetick
self.spread_last_mid = self.spread_mid_array[-1]
# 计算信号并发单
if self.get_pos(self.leg1_symbol) == 0 and self.get_pos(
self.leg2_symbol) == 0:
if self.spread_last_mid >= self.ema_up:
# 做空价差
tick_time = tick.datetime
code_time = datetime.now()
self.write_log(
f"code{code_time} tick{tick_time} in on_tick:{self.spread_last_mid}>={self.ema_up},价差开空头"
)
#print(f"code{code_time} tick{tick_time} in on_tick:{self.spread_last_mid}>={self.ema_up},价差开空头")
vt_orderids = self.short(self.leg1_symbol, self.leg1_last_ask,
self.arb_size)
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
vt_orderids = self.buy(self.leg2_symbol, self.leg2_last_bid,
self.arb_size)
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
self.try_open = 1
if self.spread_last_mid <= self.ema_down:
# 做多价差
tick_time = tick.datetime
code_time = datetime.now()
self.write_log(
f"code{code_time} tick{tick_time} in on_tick:{self.spread_last_mid}<={self.ema_down},价差开多头"
)
#print(f"code{code_time} tick{tick_time} in on_tick:{self.spread_last_mid}<={self.ema_down},价差开多头")
vt_orderids = self.buy(self.leg1_symbol, self.leg1_last_bid,
self.arb_size)
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
vt_orderids = self.short(self.leg2_symbol, self.leg2_last_ask,
self.arb_size)
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
self.try_open = 1
if self.get_pos(self.leg1_symbol) > 0 and self.get_pos(
self.leg2_symbol) < 0:
if self.spread_last_mid >= self.ema_mid:
# 反手做空价差
tick_time = tick.datetime
code_time = datetime.now()
self.write_log(
f"code{code_time} tick{tick_time} in on_tick:{self.spread_last_mid}>={self.ema_mid},价差多头退出"
)
#print(f"code{code_time} tick{tick_time} in on_tick:{self.spread_last_mid}>={self.ema_mid},价差多头退出")
vt_orderids = self.sell(self.leg1_symbol, self.leg1_last_ask,
self.arb_size)
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
vt_orderids = self.cover(self.leg2_symbol, self.leg2_last_bid,
self.arb_size)
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
if self.get_pos(self.leg1_symbol) < 0 and self.get_pos(
self.leg2_symbol) > 0:
if self.spread_last_mid <= self.ema_mid:
# 反手做多价差
tick_time = tick.datetime
code_time = datetime.now()
self.write_log(
f"code{code_time} tick{tick_time} in on_tick:{self.spread_last_mid}<={self.ema_mid},价差空头退出"
)
#print(f"code{code_time} tick{tick_time} in on_tick:{self.spread_last_mid}<={self.ema_mid},价差空头退出")
vt_orderids = self.cover(self.leg1_symbol, self.leg1_last_bid,
self.arb_size)
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
vt_orderids = self.sell(self.leg2_symbol, self.leg2_last_ask,
self.arb_size)
if vt_orderids:
for vt_orderid in vt_orderids:
self.active_orderids.add(vt_orderid)
def calc(marketItem):
# 生配列をnp DataFrameに変換
df_ohclv = pd.DataFrame(
marketItem,
columns=['Timestamp', 'Open', 'High', 'Low', 'Close', 'Volume'])
# 欠損値の削除
df_ohclv = df_ohclv.dropna()
# PlotlyとTA-libでテクニカル分析チャートを描く
# https://akatak.hatenadiary.jp/entry/2019/11/23/220836
h = np.array(df_ohclv['High'].fillna(method='ffill'))
l = np.array(df_ohclv['Low'].fillna(method='ffill'))
c = np.array(df_ohclv['Close'].fillna(method='ffill'))
# Simple Moving Average 5
df_ohclv['Sma5'] = ta.SMA(c, timeperiod=5)
# Simple Moving Average 10
df_ohclv['Sma10'] = ta.SMA(c, timeperiod=10)
# Simple Moving Average 20
df_ohclv['Sma20'] = ta.SMA(c, timeperiod=20)
# Simple Moving Average 60
df_ohclv['Sma60'] = ta.SMA(c, timeperiod=60)
# Exponential Moving Average 5
df_ohclv['Ema5'] = ta.EMA(c, timeperiod=5)
# Exponential Moving Average 20
df_ohclv['Ema20'] = ta.EMA(c, timeperiod=20)
# Exponential Moving Average 40
df_ohclv['Ema40'] = ta.EMA(c, timeperiod=40)
# 標準偏差
df_ohclv['StdDev'] = ta.STDDEV(c, timeperiod=26)
# ADX(Average Directional Movement Index 平均方向性指数)
df_ohclv['Adx'] = ta.ADX(h, l, c, timeperiod=14)
# Bollinger Bands
ary_uband, ary_mband, ary_lband = ta.BBANDS(c,
timeperiod=21,
nbdevup=0.75,
nbdevdn=0.75,
matype=0)
df_ohclv['BbH'] = ary_uband
df_ohclv['BbM'] = ary_mband
df_ohclv['BbL'] = ary_lband
# ATR(Average True Range)
df_ohclv['ATR'] = ta.ATR(h, l, c, timeperiod=14)
# SIC (期間中の最大と最小)
df_ohclv["SicH"] = pd.Series(c).rolling(window=20).max()
df_ohclv["SicL"] = pd.Series(c).rolling(window=20).min()
# SAR (Stop and Reverse)
df_ohclv["SarH"] = df_ohclv["SicH"] - df_ohclv['ATR'] * 3.3
df_ohclv["SarL"] = df_ohclv["SicL"] + df_ohclv['ATR'] * 3.3
# SAR Trend
ary_SarH = np.array(df_ohclv["SarH"])
ary_SarL = np.array(df_ohclv["SarL"])
ary_SarT = []
ary_SarTp = []
ary_SarTm = []
ary_Sar = []
# 標準偏差ボラティリティトレード
ary_StdDev = np.array(df_ohclv["StdDev"])
ary_Adx = np.array(df_ohclv['Adx'])
ary_SdBlaT = []
ary_SdBlaTp = []
ary_SdBlaTm = []
# 移動平均大循環分析 & 乖離率
ary_SMA5 = np.array(df_ohclv["Sma5"])
ary_SMA20 = np.array(df_ohclv["Sma20"])
ary_SMA60 = np.array(df_ohclv["Sma60"])
ary_MACycle = []
ary_MACycleT = []
ary_MACycleT1 = []
ary_MACycleT2 = []
ary_MACycleT3 = []
ary_MACycleT4 = []
ary_MACycleT5 = []
ary_MACycleT6 = []
ary_MADiff_5_60 = []
# Trend Balance Point
# モメンタムファクター 当日終値とTBP_K日前の終値の差分
ary_mf = pd.Series(df_ohclv.Close) - pd.Series(df_ohclv.Close).shift(2)
# TBP 上昇トレンド転換値(上昇トレンド、TBP(当日) ＝ 前々日終値 ＋ (前日MF、前々日MFの小さい方）)
ary_Tbpp = pd.Series(df_ohclv.Close).shift(2) + np.min(
[ary_mf.shift(1), ary_mf.shift(2)], axis=0)
# TBP 下降トレンド転換値(下落トレンド、TBP(当日) ＝ 前々日終値 ＋ (前日MF、前々日MFの大きい方）)
ary_Tbpm = pd.Series(df_ohclv.Close).shift(2) + np.max(
[ary_mf.shift(1), ary_mf.shift(2)], axis=0)
# np.max([df_ohclv.High - df_ohclv.Low,
# # (pd.Series(df_ohclv.Close).shift(1) - df_ohclv.High).abs(),
# # (pd.Series(df_ohclv.Close).shift(1) - df_ohclv.Low).abs()],
# # axis=0)
# 自分でループしないと計算できないヤツラ
for idx in range(df_ohclv.shape[0]):
if idx == 0:
ary_SarT.append(0)
ary_SarTp.append(0)
ary_SarTm.append(0)
ary_Sar.append(0)
ary_SdBlaT.append(0)
ary_SdBlaTp.append(0)
ary_SdBlaTm.append(0)
ary_MACycle.append(0)
ary_MACycleT.append(0)
ary_MACycleT1.append(0)
ary_MACycleT2.append(0)
ary_MACycleT3.append(0)
ary_MACycleT4.append(0)
ary_MACycleT5.append(0)
ary_MACycleT6.append(0)
ary_MADiff_5_60.append(0)
continue
# SAR (Stop and Reverse)
# SarLをCloseが上抜け(Golden Cross)
if ary_SarL[idx - 1] > c[idx - 1] and \
ary_SarL[idx] <= c[idx]:
ary_SarT.append(1)
ary_SarTp.append(1)
ary_SarTm.append(0)
ary_Sar.append(ary_SarH[idx])
# SarHをCloseが下抜け(Golden Cross)
elif ary_SarH[idx - 1] < c[idx - 1] and \
ary_SarH[idx] >= c[idx]:
ary_SarT.append(-1)
ary_SarTp.append(0)
ary_SarTm.append(1)
ary_Sar.append(ary_SarL[idx])
else:
# 前日のトレンドを保持
ary_SarT.append(ary_SarT[idx - 1])
ary_SarTp.append(ary_SarTp[idx - 1])
ary_SarTm.append(ary_SarTm[idx - 1])
if ary_SarT[idx] == 1:
ary_Sar.append(ary_SarH[idx])
elif ary_SarT[idx] == -1:
ary_Sar.append(ary_SarL[idx])
else:
ary_Sar.append(0)
# 標準偏差ボラティリティトレードモデル
if ary_StdDev[idx] > ary_StdDev[idx - 1] and \
ary_Adx[idx] > ary_Adx[idx - 1] and \
c[idx] >= ary_uband[idx]:
ary_SdBlaT.append(1)
ary_SdBlaTp.append(1)
ary_SdBlaTm.append(0)
elif ary_StdDev[idx] > ary_StdDev[idx - 1] and \
ary_Adx[idx] > ary_Adx[idx - 1] and \
c[idx] <= ary_lband[idx]:
ary_SdBlaT.append(-1)
ary_SdBlaTp.append(0)
ary_SdBlaTm.append(1)
elif ary_SdBlaT[idx - 1] == 1 and \
c[idx] >= ary_uband[idx]:
ary_SdBlaT.append(1)
ary_SdBlaTp.append(1)
ary_SdBlaTm.append(0)
elif ary_SdBlaT[idx - 1] == -1 and \
c[idx] <= ary_lband[idx]:
ary_SdBlaT.append(-1)
ary_SdBlaTp.append(0)
ary_SdBlaTm.append(1)
else:
ary_SdBlaT.append(0)
ary_SdBlaTp.append(0)
ary_SdBlaTm.append(0)
# 移動平均大循環分析 & 乖離率
if ary_SMA5[idx] >= ary_SMA20[idx] >= ary_SMA60[idx]:
ary_MACycle.append("ST1")
ary_MACycleT.append(1)
ary_MACycleT1.append(1)
ary_MACycleT2.append(0)
ary_MACycleT3.append(0)
ary_MACycleT4.append(0)
ary_MACycleT5.append(0)
ary_MACycleT6.append(0)
ary_MADiff_5_60.append(ary_SMA5[idx] / ary_SMA60[idx] * 100)
elif ary_SMA20[idx] >= ary_SMA5[idx] >= ary_SMA60[idx]:
ary_MACycle.append("ST2")
ary_MACycleT.append(0)
ary_MACycleT1.append(0)
ary_MACycleT2.append(1)
ary_MACycleT3.append(0)
ary_MACycleT4.append(0)
ary_MACycleT5.append(0)
ary_MACycleT6.append(0)
ary_MADiff_5_60.append(ary_SMA5[idx] / ary_SMA60[idx] * 100)
elif ary_SMA20[idx] >= ary_SMA60[idx] >= ary_SMA5[idx]:
ary_MACycle.append("ST3")
ary_MACycleT.append(0)
ary_MACycleT1.append(0)
ary_MACycleT2.append(0)
ary_MACycleT3.append(1)
ary_MACycleT4.append(0)
ary_MACycleT5.append(0)
ary_MACycleT6.append(0)
ary_MADiff_5_60.append(ary_SMA5[idx] / ary_SMA60[idx] * 100)
elif ary_SMA60[idx] >= ary_SMA20[idx] >= ary_SMA5[idx]:
ary_MACycle.append("ST4")
ary_MACycleT.append(-1)
ary_MACycleT1.append(0)
ary_MACycleT2.append(0)
ary_MACycleT3.append(0)
ary_MACycleT4.append(1)
ary_MACycleT5.append(0)
ary_MACycleT6.append(0)
ary_MADiff_5_60.append(ary_SMA5[idx] / ary_SMA60[idx] * 100)
elif ary_SMA60[idx] >= ary_SMA5[idx] >= ary_SMA20[idx]:
ary_MACycle.append("ST5")
ary_MACycleT.append(0)
ary_MACycleT1.append(0)
ary_MACycleT2.append(0)
ary_MACycleT3.append(0)
ary_MACycleT4.append(0)
ary_MACycleT5.append(1)
ary_MACycleT6.append(0)
ary_MADiff_5_60.append(ary_SMA5[idx] / ary_SMA60[idx] * 100)
elif ary_SMA5[idx] >= ary_SMA60[idx] >= ary_SMA20[idx]:
ary_MACycle.append("ST6")
ary_MACycleT.append(0)
ary_MACycleT1.append(0)
ary_MACycleT2.append(0)
ary_MACycleT3.append(0)
ary_MACycleT4.append(0)
ary_MACycleT5.append(0)
ary_MACycleT6.append(1)
ary_MADiff_5_60.append(ary_SMA5[idx] / ary_SMA60[idx] * 100)
else:
ary_MACycle.append("None")
ary_MACycleT.append(0)
ary_MACycleT1.append(0)
ary_MACycleT2.append(0)
ary_MACycleT3.append(0)
ary_MACycleT4.append(0)
ary_MACycleT5.append(0)
ary_MACycleT6.append(0)
ary_MADiff_5_60.append(0)
df_ohclv["SarT"] = ary_SarT
df_ohclv["SarTp"] = ary_SarTp
df_ohclv["SarTm"] = ary_SarTm
df_ohclv["Sar"] = ary_Sar
df_ohclv["SdBlaT"] = ary_SdBlaT
df_ohclv["SdBlaTp"] = ary_SdBlaTp
df_ohclv["SdBlaTm"] = ary_SdBlaTm
df_ohclv["MACycle"] = ary_MACycle
df_ohclv["MACycleT"] = ary_MACycleT
df_ohclv["MACycleT1"] = ary_MACycleT1
df_ohclv["MACycleT2"] = ary_MACycleT2
df_ohclv["MACycleT3"] = ary_MACycleT3
df_ohclv["MACycleT4"] = ary_MACycleT4
df_ohclv["MACycleT5"] = ary_MACycleT5
df_ohclv["MACycleT6"] = ary_MACycleT6
df_ohclv["MADiff_5_60"] = ary_MADiff_5_60
df_ohclv["TbpMf"] = ary_mf
df_ohclv["TbpP"] = ary_Tbpp
df_ohclv["TbpM"] = ary_Tbpm
# # Simple Moving Average 5
# df_ohclv['Sma5'] = pd.Series(df_ohclv.Close).rolling(window=5).mean()
# # Simple Moving Average 20
# df_ohclv['Sma20'] = pd.Series(df_ohclv.Close).rolling(window=20).mean()
# # Simple Moving Average 60
# df_ohclv['Sma60'] = pd.Series(df_ohclv.Close).rolling(window=60).mean()
#
# # Exponential Moving Average 5
# df_ohclv['Ema5'] = pd.Series(df_ohclv.Close).ewm(span=5).mean()
# # Exponential Moving Average 20
# df_ohclv['Ema20'] = pd.Series(df_ohclv.Close).ewm(span=20).mean()
# # Exponential Moving Average 40
# df_ohclv['Ema40'] = pd.Series(df_ohclv.Close).ewm(span=40).mean()
#
# # True Range
# df_ohclv['Tr'] = np.max([df_ohclv.High - df_ohclv.Low,
# (pd.Series(df_ohclv.Close).shift(1) - df_ohclv.High).abs(),
# (pd.Series(df_ohclv.Close).shift(1) - df_ohclv.Low).abs()],
# axis=0)
# # Average True Range
# df_ohclv["Atr"] = pd.Series(df_ohclv.Tr).rolling(window=7).mean()
#
# # SIC (期間中の最大と最小)
# df_ohclv["SicH"] = pd.Series(df_ohclv.Close).rolling(window=20).max()
# df_ohclv["SicL"] = pd.Series(df_ohclv.Close).rolling(window=20).min()
#
# # Wilder Volatility System Trend
# df_ohclv["WVST"] = "None"
#
#
#
# # Stop And Reverse (SAR)
return df_ohclv
def ema(dataframe, period, field='close'):
import talib.abstract as ta
return ta.EMA(dataframe, timeperiod=period, price=field)
def data_construct(DataFrame, lookUp, predictionWindow, pairName):
'''function to construct the features from the inspection window and to create the supervised x,y pairs for training.
Parameters
----------
DataFrame : dataFrame
LookUp : int
predictionWindow : int
pairName : str
Returns
-------
output : dict
a dict containing inputs matrix, targets matrix, raw inputs and mapping dict for features
'''
# fetch data for indicators calculations
openPrice = DataFrame.o.values.astype("double")
closePrice = DataFrame.c.values.astype("double")
highPrice = DataFrame.h.values.astype("double")
lowPrice = DataFrame.l.values.astype("double")
volume = DataFrame.volume.values.astype("double")
# calculate technical indicators values
simple_ma_slow = ta.SMA(closePrice, 30) # slow moving average
simple_ma_fast = ta.SMA(closePrice, 15) # fast moving average
exp_ma_slow = ta.EMA(closePrice, 20) # slow exp moving average
exp_ma_fast = ta.EMA(closePrice, 10) # fast exp moving average
bbands = ta.BBANDS(closePrice, timeperiod=15) # calculate bollinger bands
deltaBands = (bbands[0] - bbands[2]
) / bbands[2] # deltas between bands vector (bollinger)
macd_s1, macd_s2, macd_hist = ta.MACD(
closePrice) # MACD values calculation
sar = ta.SAR(highPrice, lowPrice) # prabolic SAR
stochK, stochD = ta.STOCH(highPrice, lowPrice,
closePrice) # stochastic calculations
rsi = ta.RSI(closePrice, timeperiod=15) # RSI indicator
adx = ta.ADX(highPrice, lowPrice, closePrice,
timeperiod=15) # ADX indicator
mfi = ta.MFI(highPrice, lowPrice, closePrice, volume,
timeperiod=15) # money flow index
# calculate statistical indicators values
beta = ta.BETA(highPrice, lowPrice, timeperiod=5) # beta from CAPM model
slope = ta.LINEARREG_ANGLE(
closePrice,
timeperiod=5) # slope for fitting linera reg. to the last x points
# calculate candle indicators values
spinTop = ta.CDLSPINNINGTOP(openPrice, highPrice, lowPrice, closePrice)
doji = ta.CDLDOJI(openPrice, highPrice, lowPrice, closePrice)
dojiStar = ta.CDLDOJISTAR(openPrice, highPrice, lowPrice, closePrice)
marubozu = ta.CDLMARUBOZU(openPrice, highPrice, lowPrice, closePrice)
hammer = ta.CDLHAMMER(openPrice, highPrice, lowPrice, closePrice)
invHammer = ta.CDLINVERTEDHAMMER(openPrice, highPrice, lowPrice,
closePrice)
hangingMan = ta.CDLHANGINGMAN(openPrice, highPrice, lowPrice, closePrice)
shootingStar = ta.CDLSHOOTINGSTAR(openPrice, highPrice, lowPrice,
closePrice)
engulfing = ta.CDLENGULFING(openPrice, highPrice, lowPrice, closePrice)
morningStar = ta.CDLMORNINGSTAR(openPrice, highPrice, lowPrice, closePrice)
eveningStar = ta.CDLEVENINGSTAR(openPrice, highPrice, lowPrice, closePrice)
whiteSoldier = ta.CDL3WHITESOLDIERS(openPrice, highPrice, lowPrice,
closePrice)
blackCrow = ta.CDL3BLACKCROWS(openPrice, highPrice, lowPrice, closePrice)
insideThree = ta.CDL3INSIDE(openPrice, highPrice, lowPrice, closePrice)
# prepare the final matrix
'''
matrix configurations ::> [o,c,h,l,ma_slow,ma_fast,exp_slow,exp_fast,
deltaBands,macd_s1,macd_s2,sar,stochK,
stochD,rsi,adx,mfi,beta,slope,spinTop,doji,dojiStar,
marubozu,hammer,invHammer,hangingMan,shootingStar,engulfing,
morningStar,eveningStar,whiteSoldier,blackCrow,insideThree]
a 33 features matrix in total
'''
DataMatrix = np.column_stack(
(openPrice, closePrice, highPrice, lowPrice, simple_ma_slow,
simple_ma_fast, exp_ma_slow, exp_ma_fast, deltaBands, macd_s1,
macd_s2, sar, stochK, stochD, rsi, adx, mfi, beta, slope, spinTop,
doji, dojiStar, marubozu, hammer, invHammer, hangingMan, shootingStar,
engulfing, morningStar, eveningStar, whiteSoldier, blackCrow,
insideThree))
# remove undifined values
DataMatrix = DataMatrix[~np.isnan(DataMatrix).any(
axis=1)] # remove all raws containing nan values
# define number of windows to analyze
framesCount = DataMatrix.shape[0] - (
lookUp +
predictionWindow) + 1 # 1D convolution outputsize = ceil[((n-f)/s)+1]
# define input/output arrays container
rawInputs = {}
inputsOpen = np.zeros((framesCount, lookUp))
inputsClose = np.zeros((framesCount, lookUp))
inputsHigh = np.zeros((framesCount, lookUp))
inputsLow = np.zeros((framesCount, lookUp))
inputs = np.zeros((framesCount, 62))
outputs = np.zeros((framesCount, 1))
# main loop and data
for i in range(framesCount):
mainFrame = DataMatrix[i:i + lookUp + predictionWindow, :]
window = np.array_split(mainFrame, [lookUp])[0]
windowForecast = np.array_split(mainFrame, [lookUp])[1]
'''
window configurations ::>
[0:o,1:c,2:h,3:l,4:ma_slow,5:ma_fast,6:exp_slow,7:exp_fast,
8:deltaBands,9:macd_slow,10:macd_fast,11:sar,12:stochK,
13:stochD,14:rsi,15:adx,16:mfi,17:beta,18:slope,19:spinTop,20:doji,21:dojiStar,
22:marubozu,23:hammer,24:invHammer,25:hangingMan,26:shootingStar,27:engulfing,
28:morningStar,29:eveningStar,30:whiteSoldier,31:blackCrow,32:insideThree]
'''
#sma features detection
ma_slow = window[:, 4]
ma_fast = window[:, 5]
uptrend_cross = ma_fast > ma_slow
uptrend_cross = np.concatenate(
(np.array([False]),
(uptrend_cross[:-1] <
uptrend_cross[1:]))) # check the false->true transition
try:
uptrend_cross_location = np.where(uptrend_cross == True)[0][
-1] # latest uptrend cross_over location
except:
uptrend_cross_location = -1
downtrend_cross = ma_slow > ma_fast
downtrend_cross = np.concatenate(
(np.array([False]),
(downtrend_cross[:-1] <
downtrend_cross[1:]))) # check the false->true transition
try:
downtrend_cross_location = np.where(downtrend_cross == True)[0][
-1] # latest downtrend cross_over location
except:
downtrend_cross_location = -1
if (uptrend_cross_location >
downtrend_cross_location): # latest cross is an uptrend
sma_latest_crossover = 1 # uptrend sign
sma_location_of_latest_crossover = uptrend_cross_location
alpha_1 = (math.atan(ma_slow[uptrend_cross_location] -
ma_slow[uptrend_cross_location - 1])) * (
180 / math.pi)
alpha_2 = (math.atan(ma_fast[uptrend_cross_location] -
ma_fast[uptrend_cross_location - 1])) * (
180 / math.pi)
sma_latest_crossover_angle = alpha_1 + alpha_2
elif (downtrend_cross_location >
uptrend_cross_location): # latest cross is a downtrend
sma_latest_crossover = -1 # downtrend sign
sma_location_of_latest_crossover = downtrend_cross_location
alpha_1 = (math.atan(ma_slow[downtrend_cross_location] -
ma_slow[downtrend_cross_location - 1])) * (
180 / math.pi)
alpha_2 = (math.atan(ma_fast[downtrend_cross_location] -
ma_fast[downtrend_cross_location - 1])) * (
180 / math.pi)
sma_latest_crossover_angle = alpha_1 + alpha_2
else: # no cross in the given window
sma_latest_crossover = 0 # no sign
sma_location_of_latest_crossover = -1
sma_latest_crossover_angle = 0
up_count = np.sum(ma_fast > ma_slow)
down_count = np.sum(ma_slow > ma_fast)
if (up_count > down_count):
sma_dominant_type_fast_slow = 1
elif (down_count > up_count):
sma_dominant_type_fast_slow = -1
else:
sma_dominant_type_fast_slow = 0
#ema features detection
exp_slow = window[:, 6]
exp_fast = window[:, 7]
uptrend_cross = exp_fast > exp_slow
uptrend_cross = np.concatenate(
(np.array([False]),
(uptrend_cross[:-1] <
uptrend_cross[1:]))) # check the false->true transition
try:
uptrend_cross_location = np.where(uptrend_cross == True)[0][
-1] # latest uptrend cross_over location
except:
uptrend_cross_location = -1
downtrend_cross = exp_slow > exp_fast
downtrend_cross = np.concatenate(
(np.array([False]),
(downtrend_cross[:-1] <
downtrend_cross[1:]))) # check the false->true transition
try:
downtrend_cross_location = np.where(downtrend_cross == True)[0][
-1] # latest downtrend cross_over location
except:
downtrend_cross_location = -1
if (uptrend_cross_location >
downtrend_cross_location): # latest cross is an uptrend
ema_latest_crossover = 1 # uptrend sign
ema_location_of_latest_crossover = uptrend_cross_location
alpha_1 = (math.atan(exp_slow[uptrend_cross_location] -
exp_slow[uptrend_cross_location - 1])) * (
180 / math.pi)
alpha_2 = (math.atan(exp_fast[uptrend_cross_location] -
exp_fast[uptrend_cross_location - 1])) * (
180 / math.pi)
ema_latest_crossover_angle = alpha_1 + alpha_2
elif (downtrend_cross_location >
uptrend_cross_location): # latest cross is a downtrend
ema_latest_crossover = -1 # downtrend sign
ema_location_of_latest_crossover = downtrend_cross_location
alpha_1 = (math.atan(exp_slow[downtrend_cross_location] -
exp_slow[downtrend_cross_location - 1])) * (
180 / math.pi)
alpha_2 = (math.atan(exp_fast[downtrend_cross_location] -
exp_fast[downtrend_cross_location - 1])) * (
180 / math.pi)
ema_latest_crossover_angle = alpha_1 + alpha_2
else: # no cross in the given window
ema_latest_crossover = 0 # no sign
ema_location_of_latest_crossover = -1
ema_latest_crossover_angle = 0
up_count = np.sum(exp_fast > exp_slow)
down_count = np.sum(exp_slow > exp_fast)
if (up_count > down_count):
ema_dominant_type_fast_slow = 1
elif (down_count > up_count):
ema_dominant_type_fast_slow = -1
else:
ema_dominant_type_fast_slow = 0
# B.Bands features detection
deltaBands = window[:, 8]
deltaBands_mean = np.mean(deltaBands)
deltaBands_std = np.std(deltaBands)
deltaBands_maximum_mean = np.amax(deltaBands) / deltaBands_mean
deltaBands_maximum_location = np.where(
deltaBands == np.amax(deltaBands))[0][-1] # location of maximum
deltaBands_minimum_mean = np.amin(deltaBands) / deltaBands_mean
deltaBands_minimum_location = np.where(
deltaBands == np.amin(deltaBands))[0][-1] # location of maximum
# macd features detection
macd_slow = window[:, 9]
macd_fast = window[:, 10]
uptrend_cross = macd_fast > macd_slow
uptrend_cross = np.concatenate(
(np.array([False]),
(uptrend_cross[:-1] <
uptrend_cross[1:]))) # check the false->true transition
try:
uptrend_cross_location = np.where(uptrend_cross == True)[0][
-1] # latest uptrend cross_over location
except:
uptrend_cross_location = -1
downtrend_cross = macd_slow > macd_fast
downtrend_cross = np.concatenate(
(np.array([False]),
(downtrend_cross[:-1] <
downtrend_cross[1:]))) # check the false->true transition
try:
downtrend_cross_location = np.where(downtrend_cross == True)[0][
-1] # latest downtrend cross_over location
except:
downtrend_cross_location = -1
if (uptrend_cross_location >
downtrend_cross_location): # latest cross is an uptrend
macd_latest_crossover = 1 # uptrend sign
macd_location_of_latest_crossover = uptrend_cross_location
alpha_1 = (math.atan(macd_slow[uptrend_cross_location] -
macd_slow[uptrend_cross_location - 1])) * (
180 / math.pi)
alpha_2 = (math.atan(macd_fast[uptrend_cross_location] -
macd_fast[uptrend_cross_location - 1])) * (
180 / math.pi)
macd_latest_crossover_angle = alpha_1 + alpha_2
elif (downtrend_cross_location >
uptrend_cross_location): # latest cross is a downtrend
macd_latest_crossover = -1 # downtrend sign
macd_location_of_latest_crossover = downtrend_cross_location
alpha_1 = (math.atan(macd_slow[downtrend_cross_location] -
macd_slow[downtrend_cross_location - 1])) * (
180 / math.pi)
alpha_2 = (math.atan(macd_fast[downtrend_cross_location] -
macd_fast[downtrend_cross_location - 1])) * (
180 / math.pi)
macd_latest_crossover_angle = alpha_1 + alpha_2
else: # no cross in the given window
macd_latest_crossover = 0 # no sign
macd_location_of_latest_crossover = -1
macd_latest_crossover_angle = 0
up_count = np.sum(macd_fast > macd_slow)
down_count = np.sum(macd_slow > macd_fast)
if (up_count > down_count):
macd_dominant_type_fast_slow = 1
elif (down_count > up_count):
macd_dominant_type_fast_slow = -1
else:
macd_dominant_type_fast_slow = 0
# sar features detection
average_price = (window[:, 0] + window[:, 1] + window[:, 2] +
window[:, 3]) / 4
sar = window[:, 11]
uptrend = sar < average_price
uptrend = np.concatenate(
(np.array([False]),
(uptrend[:-1] < uptrend[1:]))) # check the false->true transition
try:
uptrend_location = np.where(
uptrend == True)[0][-1] # latest uptrend location
except:
uptrend_location = -1
downtrend = sar > average_price
downtrend = np.concatenate(
(np.array([False]),
(downtrend[:-1] <
downtrend[1:]))) # check the false->true transition
try:
downtrend_location = np.where(
downtrend == True)[0][-1] # latest downtrend location
except:
downtrend_location = -1
if (uptrend_location >
downtrend_location): # latest signal is an uptrend
sar_latest_shiftPoint = 1
sar_latest_shiftPoint_location = uptrend_location
elif (downtrend_location >
uptrend_location): # latest signal is a downtrend
sar_latest_shiftPoint = -1
sar_latest_shiftPoint_location = downtrend_location
else: # same direction along the frame under question
sar_latest_shiftPoint = 0 # no sign
sar_latest_shiftPoint_location = -1
sar_total_number_shifts = np.where(
downtrend == True)[0].shape[0] + np.where(
uptrend == True)[0].shape[0]
# stochastic(K) features detection
stochK = window[:, 12]
stochK_mean = np.mean(stochK)
stochK_std = np.std(stochK)
uptrend = stochK <= 20
uptrend = np.concatenate(
(np.array([False]),
(uptrend[:-1] < uptrend[1:]))) # check the false->true transition
try:
uptrend_location = np.where(
uptrend == True)[0][-1] # latest uptrend location
except:
uptrend_location = -1
downtrend = stochK >= 80
downtrend = np.concatenate(
(np.array([False]),
(downtrend[:-1] <
downtrend[1:]))) # check the false->true transition
try:
downtrend_location = np.where(
downtrend == True)[0][-1] # latest downtrend location
except:
downtrend_location = -1
if (uptrend_location >
downtrend_location): # latest signal is an uptrend
stochK_latest_event = 1
stochK_event_location = uptrend_location
elif (downtrend_location >
uptrend_location): # latest signal is a downtrend
stochK_latest_event = -1
stochK_event_location = downtrend_location
else: # same direction along the frame under question
stochK_latest_event = 0 # no sign
stochK_event_location = -1
# stochastic(D) features detection
stochD = window[:, 13]
stochD_mean = np.mean(stochD)
stochD_std = np.std(stochD)
uptrend = stochD <= 20
uptrend = np.concatenate(
(np.array([False]),
(uptrend[:-1] < uptrend[1:]))) # check the false->true transition
try:
uptrend_location = np.where(
uptrend == True)[0][-1] # latest uptrend location
except:
uptrend_location = -1
downtrend = stochD >= 80
downtrend = np.concatenate(
(np.array([False]),
(downtrend[:-1] <
downtrend[1:]))) # check the false->true transition
try:
downtrend_location = np.where(
downtrend == True)[0][-1] # latest downtrend location
except:
downtrend_location = -1
if (uptrend_location >
downtrend_location): # latest signal is an uptrend
stochD_latest_event = 1
stochD_event_location = uptrend_location
elif (downtrend_location >
uptrend_location): # latest signal is a downtrend
stochD_latest_event = -1
stochD_event_location = downtrend_location
else: # same direction along the frame under question
stochD_latest_event = 0 # no sign
stochD_event_location = -1
# rsi features detection
rsi = window[:, 14]
rsi_mean = np.mean(rsi)
rsi_std = np.std(rsi)
uptrend = rsi <= 30
uptrend = np.concatenate(
(np.array([False]),
(uptrend[:-1] < uptrend[1:]))) # check the false->true transition
try:
uptrend_location = np.where(
uptrend == True)[0][-1] # latest uptrend location
except:
uptrend_location = -1
downtrend = rsi >= 70
downtrend = np.concatenate(
(np.array([False]),
(downtrend[:-1] <
downtrend[1:]))) # check the false->true transition
try:
downtrend_location = np.where(
downtrend == True)[0][-1] # latest downtrend location
except:
downtrend_location = -1
if (uptrend_location >
downtrend_location): # latest signal is an uptrend
rsi_latest_event = 1
rsi_event_location = uptrend_location
elif (downtrend_location >
uptrend_location): # latest signal is a downtrend
rsi_latest_event = -1
rsi_event_location = downtrend_location
else: # same direction along the frame under question
rsi_latest_event = 0 # no sign
rsi_event_location = -1
# adx features detection
adx = window[:, 15]
adx_mean = np.mean(adx)
adx_std = np.std(adx)
splitted_array = np.array_split(adx, 2)
m0 = np.mean(splitted_array[0])
m1 = np.mean(splitted_array[1])
adx_mean_delta_bet_first_second_half = (m1 - m0) / m0
# mfi features detection
mfi = window[:, 16]
mfi_mean = np.mean(mfi)
mfi_std = np.std(mfi)
splitted_array = np.array_split(mfi, 2)
m0 = np.mean(splitted_array[0])
m1 = np.mean(splitted_array[1])
mfi_mean_delta_bet_first_second_half = (m1 - m0) / m0
# resistance levels features detection
closePrice = window[:, 1]
resLevels = argrelextrema(closePrice, np.greater, order=4)[0]
if (resLevels.shape[0] == 0):
relation_r1_close = 0
relation_r2_close = 0
relation_r3_close = 0
elif (resLevels.shape[0] == 1):
relation_r1_close = (closePrice[-1] -
closePrice[resLevels[-1]]) / closePrice[-1]
relation_r2_close = 0
relation_r3_close = 0
elif (resLevels.shape[0] == 2):
relation_r1_close = (closePrice[-1] -
closePrice[resLevels[-1]]) / closePrice[-1]
relation_r2_close = (closePrice[-1] -
closePrice[resLevels[-2]]) / closePrice[-1]
relation_r3_close = 0
else:
relation_r1_close = (closePrice[-1] -
closePrice[resLevels[-1]]) / closePrice[-1]
relation_r2_close = (closePrice[-1] -
closePrice[resLevels[-2]]) / closePrice[-1]
relation_r3_close = (closePrice[-1] -
closePrice[resLevels[-3]]) / closePrice[-1]
# support levels features detection
closePrice = window[:, 1]
supLevels = argrelextrema(closePrice, np.less, order=4)[0]
if (supLevels.shape[0] == 0):
relation_s1_close = 0
relation_s2_close = 0
relation_s3_close = 0
elif (supLevels.shape[0] == 1):
relation_s1_close = (closePrice[-1] -
closePrice[supLevels[-1]]) / closePrice[-1]
relation_s2_close = 0
relation_s3_close = 0
elif (supLevels.shape[0] == 2):
relation_s1_close = (closePrice[-1] -
closePrice[supLevels[-1]]) / closePrice[-1]
relation_s2_close = (closePrice[-1] -
closePrice[supLevels[-2]]) / closePrice[-1]
relation_s3_close = 0
else:
relation_s1_close = (closePrice[-1] -
closePrice[supLevels[-1]]) / closePrice[-1]
relation_s2_close = (closePrice[-1] -
closePrice[supLevels[-2]]) / closePrice[-1]
relation_s3_close = (closePrice[-1] -
closePrice[supLevels[-3]]) / closePrice[-1]
# slope features detection
slope = window[:, 18]
slope_mean = np.mean(slope)
# beta features detection
beta = window[:, 17]
beta_mean = np.mean(beta)
beta_std = np.std(beta)
# spinTop features detection np.sum(np.where(a==1)[0])
count100plus = np.sum(np.where(window[:, 19] == 100)[0])
count100minus = (np.sum(np.where(window[:, 19] == -100)[0])) * -1
spinTop_number_occurrence = count100plus + count100minus
# doji features detection
count100plus = np.sum(np.where(window[:, 20] == 100)[0])
count100minus = (np.sum(np.where(window[:, 20] == -100)[0])) * -1
doji_number_occurrence = count100plus + count100minus
# dojiStar features detection
count100plus = np.sum(np.where(window[:, 21] == 100)[0])
count100minus = (np.sum(np.where(window[:, 21] == -100)[0])) * -1
dojiStar_number_occurrence = count100plus + count100minus
# marubozu features detection
count100plus = np.sum(np.where(window[:, 22] == 100)[0])
count100minus = (np.sum(np.where(window[:, 22] == -100)[0])) * -1
marubozu_number_occurrence = count100plus + count100minus
# hammer features detection
count100plus = np.sum(np.where(window[:, 23] == 100)[0])
count100minus = (np.sum(np.where(window[:, 23] == -100)[0])) * -1
hammer_number_occurrence = count100plus + count100minus
# invHammer features detection
count100plus = np.sum(np.where(window[:, 24] == 100)[0])
count100minus = (np.sum(np.where(window[:, 24] == -100)[0])) * -1
invHammer_number_occurrence = count100plus + count100minus
# hangingMan features detection
count100plus = np.sum(np.where(window[:, 25] == 100)[0])
count100minus = (np.sum(np.where(window[:, 25] == -100)[0])) * -1
hangingMan_number_occurrence = count100plus + count100minus
# shootingStar features detection
count100plus = np.sum(np.where(window[:, 26] == 100)[0])
count100minus = (np.sum(np.where(window[:, 26] == -100)[0])) * -1
shootingStar_number_occurrence = count100plus + count100minus
# engulfing features detection
count100plus = np.sum(np.where(window[:, 27] == 100)[0])
count100minus = (np.sum(np.where(window[:, 27] == -100)[0])) * -1
engulfing_number_occurrence = count100plus + count100minus
# morningStar features detection
count100plus = np.sum(np.where(window[:, 28] == 100)[0])
count100minus = (np.sum(np.where(window[:, 28] == -100)[0])) * -1
morningStar_number_occurrence = count100plus + count100minus
# eveningStar features detection
count100plus = np.sum(np.where(window[:, 29] == 100)[0])
count100minus = (np.sum(np.where(window[:, 29] == -100)[0])) * -1
eveningStar_number_occurrence = count100plus + count100minus
# whiteSoldier features detection
count100plus = np.sum(np.where(window[:, 30] == 100)[0])
count100minus = (np.sum(np.where(window[:, 30] == -100)[0])) * -1
whiteSoldier_number_occurrence = count100plus + count100minus
# blackCrow features detection
count100plus = np.sum(np.where(window[:, 31] == 100)[0])
count100minus = (np.sum(np.where(window[:, 31] == -100)[0])) * -1
blackCrow_number_occurrence = count100plus + count100minus
# insideThree features detection
count100plus = np.sum(np.where(window[:, 32] == 100)[0])
count100minus = (np.sum(np.where(window[:, 32] == -100)[0])) * -1
insideThree_number_occurrence = count100plus + count100minus
# fill the inputs matrix
inputs[i, 0] = sma_latest_crossover
inputs[i, 1] = sma_location_of_latest_crossover
inputs[i, 2] = sma_latest_crossover_angle
inputs[i, 3] = sma_dominant_type_fast_slow
inputs[i, 4] = ema_latest_crossover
inputs[i, 5] = ema_location_of_latest_crossover
inputs[i, 6] = ema_latest_crossover_angle
inputs[i, 7] = ema_dominant_type_fast_slow
inputs[i, 8] = deltaBands_mean
inputs[i, 9] = deltaBands_std
inputs[i, 10] = deltaBands_maximum_mean
inputs[i, 11] = deltaBands_maximum_location
inputs[i, 12] = deltaBands_minimum_mean
inputs[i, 13] = deltaBands_minimum_location
inputs[i, 14] = macd_latest_crossover
inputs[i, 15] = macd_location_of_latest_crossover
inputs[i, 16] = macd_latest_crossover_angle
inputs[i, 17] = macd_dominant_type_fast_slow
inputs[i, 18] = sar_latest_shiftPoint
inputs[i, 19] = sar_latest_shiftPoint_location
inputs[i, 20] = sar_total_number_shifts
inputs[i, 21] = stochK_mean
inputs[i, 22] = stochK_std
inputs[i, 23] = stochK_latest_event
inputs[i, 24] = stochK_event_location
inputs[i, 25] = stochD_mean
inputs[i, 26] = stochD_std
inputs[i, 27] = stochD_latest_event
inputs[i, 28] = stochD_event_location
inputs[i, 29] = rsi_mean
inputs[i, 30] = rsi_std
inputs[i, 31] = rsi_latest_event
inputs[i, 32] = rsi_event_location
inputs[i, 33] = adx_mean
inputs[i, 34] = adx_std
inputs[i, 35] = adx_mean_delta_bet_first_second_half
inputs[i, 36] = mfi_mean
inputs[i, 37] = mfi_std
inputs[i, 38] = mfi_mean_delta_bet_first_second_half
inputs[i, 39] = relation_r1_close
inputs[i, 40] = relation_r2_close
inputs[i, 41] = relation_r3_close
inputs[i, 42] = relation_s1_close
inputs[i, 43] = relation_s2_close
inputs[i, 44] = relation_s3_close
inputs[i, 45] = slope_mean
inputs[i, 46] = beta_mean
inputs[i, 47] = beta_std
inputs[i, 48] = spinTop_number_occurrence
inputs[i, 49] = doji_number_occurrence
inputs[i, 50] = dojiStar_number_occurrence
inputs[i, 51] = marubozu_number_occurrence
inputs[i, 52] = hammer_number_occurrence
inputs[i, 53] = invHammer_number_occurrence
inputs[i, 54] = hangingMan_number_occurrence
inputs[i, 55] = shootingStar_number_occurrence
inputs[i, 56] = engulfing_number_occurrence
inputs[i, 57] = morningStar_number_occurrence
inputs[i, 58] = eveningStar_number_occurrence
inputs[i, 59] = whiteSoldier_number_occurrence
inputs[i, 60] = blackCrow_number_occurrence
inputs[i, 61] = insideThree_number_occurrence
# fill raw inputs matrices
inputsOpen[i, :] = window[:, 0].reshape(1, lookUp)
inputsClose[i, :] = window[:, 1].reshape(1, lookUp)
inputsHigh[i, :] = window[:, 2].reshape(1, lookUp)
inputsLow[i, :] = window[:, 3].reshape(1, lookUp)
# fill the output matrix
futureClose = windowForecast[:, 1]
if (pairName == "USD_JPY"):
outputs[
i, 0] = (futureClose[-1] - futureClose[0]
) / 0.01 # one pip = 0.01 for any pair containing JPY
else:
outputs[i, 0] = (futureClose[-1] - futureClose[0]
) / 0.0001 # one pip = 0.0001 for this pairs
# create mapping dict.
mappingDict = {
"sma_latest_crossover": 0,
"sma_location_of_latest_crossover": 1,
"sma_latest_crossover_angle": 2,
"sma_dominant_type_fast_slow": 3,
"ema_latest_crossover": 4,
"ema_location_of_latest_crossover": 5,
"ema_latest_crossover_angle": 6,
"ema_dominant_type_fast_slow": 7,
"deltaBands_mean": 8,
"deltaBands_std": 9,
"deltaBands_maximum_mean": 10,
"deltaBands_maximum_location": 11,
"deltaBands_minimum_mean": 12,
"deltaBands_minimum_location": 13,
"macd_latest_crossover": 14,
"macd_location_of_latest_crossover": 15,
"macd_latest_crossover_angle": 16,
"macd_dominant_type_fast_slow": 17,
"sar_latest_shiftPoint": 18,
"sar_latest_shiftPoint_location": 19,
"sar_total_number_shifts": 20,
"stochK_mean": 21,
"stochK_std": 22,
"stochK_latest_event": 23,
"stochK_event_location": 24,
"stochD_mean": 25,
"stochD_std": 26,
"stochD_latest_event": 27,
"stochD_event_location": 28,
"rsi_mean": 29,
"rsi_std": 30,
"rsi_latest_event": 31,
"rsi_event_location": 32,
"adx_mean": 33,
"adx_std": 34,
"adx_mean_delta_bet_first_second_half": 35,
"mfi_mean": 36,
"mfi_std": 37,
"mfi_mean_delta_bet_first_second_half": 38,
"relation_r1_close": 39,
"relation_r2_close": 40,
"relation_r3_close": 41,
"relation_s1_close": 42,
"relation_s2_close": 43,
"relation_s3_close": 44,
"slope_mean": 45,
"beta_mean": 46,
"beta_std": 47,
"spinTop_number_occurrence": 48,
"doji_number_occurrence": 49,
"dojiStar_number_occurrence": 50,
"marubozu_number_occurrence": 51,
"hammer_number_occurrence": 52,
"invHammer_number_occurrence": 53,
"hangingMan_number_occurrence": 54,
"shootingStar_number_occurrence": 55,
"engulfing_number_occurrence": 56,
"morningStar_number_occurrence": 57,
"eveningStar_number_occurrence": 58,
"whiteSoldier_number_occurrence": 59,
"blackCrow_number_occurrence": 60,
"insideThree_number_occurrence": 61
}
# remove undifined values from the output
refMatrix = inputs
inputs = inputs[~np.isnan(refMatrix).any(
axis=1)] # remove all raws containing nan values
outputs = outputs[~np.isnan(refMatrix).any(
axis=1)] # remove all raws containing nan values
inputsOpen = inputsOpen[~np.isnan(refMatrix).any(
axis=1)] # remove all raws containing nan values
inputsClose = inputsClose[~np.isnan(refMatrix).any(
axis=1)] # remove all raws containing nan values
inputsHigh = inputsHigh[~np.isnan(refMatrix).any(
axis=1)] # remove all raws containing nan values
inputsLow = inputsLow[~np.isnan(refMatrix).any(
axis=1)] # remove all raws containing nan values
# create raw inputs dict.
rawInputs["open"] = inputsOpen
rawInputs["close"] = inputsClose
rawInputs["high"] = inputsHigh
rawInputs["low"] = inputsLow
# return the function output
output = {
"mappingDict": mappingDict,
"rawInputs": rawInputs,
"inputFeatures": inputs,
"targets": outputs
}
return (output)
def start():
get_k_data = ts.get_k_data("sh000001", start='1990-12-19')
data = pd.DataFrame(get_k_data)
data_size = data.iloc[:, 0].size
up_list = []
down_list = []
up_down_list = []
v_list = []
per_volume_list = []
rvi_list = []
up_date_list = []
down_date_list = []
date_list = []
close_list = []
volume_list = []
qhlsr_list = []
obv_list = []
obv_rate_list = []
zzz_v_list = []
obv = 0
#昨收
y_close = 0
pre_down = 0
pre_up = 0
times = 100000000
up_volume = 1
down_volume = 1
up_times = 0
y_open = 0
y_close = 0
y_high = 0
y_low = 0
y_volume = 0
for index, row in data.iterrows():
open = row.open
close = row.close
high = row.high
low = row.low
volume = row.volume
date = row.date
date_list.append(date)
close_list.append(close)
volume_list.append(volume)
if close > y_close:
up_times = up_times + 1
up_volume = up_volume + volume
else:
down_volume = down_volume + volume
obv_rate_list.append(up_volume / down_volume)
###阻力指标
if y_volume == 0:
qhlsr = 0
else:
qhlsr = (close - y_close
) - (volume - y_volume) * (y_high - y_low) / y_volume
qhlsr_list.append(qhlsr)
zzz_v_list.append((close - y_close))
v_list.append(close - y_close)
up_list.append(times * (close - y_close) / volume)
print(times * (close - y_close) / volume)
down_list.append(times * (high - close) / volume)
up_down_list.append(up_list[index] - down_list[index])
# rvi_list.append((close-open)/(high-low));
if (high - low) == 0:
# per_volume_list.append(0);
rvi_list.append(0)
else:
# per_volume_list.append((volume/(close - y_close)));
rvi_list.append((close - open) / (high - low))
y_open = open
y_close = close
y_high = high
y_low = low
y_volume = volume
fig, axes = plt.subplots(3, 1, sharex=True, sharey=False)
# axes.plot(up_list[0:2300], 'r-')
# plt.subplots_adjust(wspace = 0, hspace = 0)
# plt.figure(figsize=(8,4))
lll = 1000
timeperiod = 20
obv_list = get_obv(data, data_size - lll, data_size)
close_ma = ta.MA(
np.array(close_list[len(close_list) - lll:len(close_list)]), 1)
# rvi_ma = ta.EMA(np.array(rvi_list[len(rvi_list) - lll:len(rvi_list)]), timeperiod)
# up_down_ma = ta.EMA(np.array(up_down_list[len(up_down_list) - lll:len(up_down_list)]), timeperiod)
rvi_ma = ta.CMO(
np.array(close_list[len(close_list) - lll:len(close_list)]),
timeperiod)
# up_down_ma = ta.ROC(np.array(close_list[len(close_list) - lll:len(close_list)]), timeperiod)
s = 50
f = 9
up_down_ema_s = ta.EMA(np.array(up_list[len(up_list) - lll:len(up_list)]),
s)
up_down_ema_f = ta.EMA(np.array(up_list[len(up_list) - lll:len(up_list)]),
f)
real = ta.LINEARREG_SLOPE(np.array(close_list[len(close_list) -
lll:len(close_list)]),
timeperiod=10)
real = ta.EMA(np.array(real), s)
up_down_ma_s = ta.MA(np.array(up_list[len(up_list) - lll:len(up_list)]), s)
up_down_ma_f = ta.MA(np.array(up_list[len(up_list) - lll:len(up_list)]), f)
v_ema_s = ta.EMA(np.array(v_list[len(v_list) - lll:len(v_list)]), s)
v_ema_f = ta.EMA(np.array(v_list[len(v_list) - lll:len(v_list)]), f)
axes[0].plot(date_list[len(date_list) - lll:len(date_list)],
close_ma,
"y--",
linewidth=1)
# axes[0].plot(date_list[len(date_list) - lll:len(date_list)],v_ema_s,"r--",linewidth=1);
# axes[1].plot(date_list[len(date_list) - lll:len(date_list)],rvi_ma,"b--",linewidth=1);
axes[1].plot(date_list[len(date_list) - lll:len(date_list)],
real,
"b--",
linewidth=1)
# axes[1].plot(date_list[len(date_list) - lll:len(date_list)],v_ema_f,"r--",linewidth=1);
axes[2].plot(date_list[len(date_list) - lll:len(date_list)],
up_down_ema_s,
"b--",
linewidth=1)
axes[2].plot(date_list[len(date_list) - lll:len(date_list)],
up_down_ema_f,
"r--",
linewidth=1)
# axes[2].plot(date_list[len(date_list) - lll:len(date_list)],,"b--",linewidth=1);
where_are_nan = np.isnan(up_down_ema_s)
where_are_inf = np.isinf(up_down_ema_s)
up_down_ema_s[where_are_nan] = 0
up_down_ema_s[where_are_inf] = 0
# print("up_list", np.mean(up_down_ema_s));
# print("v_list", np.mean(v_list));
plt.show()
def calculate_features(data: pd.DataFrame,
normalization=False,
train_data: list = None,
start=None,
end=None):
Open = data['Open'].values
High = data['High'].values
Low = data['Low'].values
Close = data['Close'].values
Volume = data['Volume'].values
data['ret'] = data['Close'].pct_change() * 100.0
data['ret_2'] = data['Close'].pct_change().shift() * 100.0
data['ret_3'] = data['Close'].pct_change().shift(2) * 100.0
data['ret_4'] = data['Close'].pct_change().shift(3) * 100.0
data['ret_5'] = data['Close'].pct_change().shift(4) * 100.0
data['ret_ratio'] = (data['ret'] / data['ret_5'] - 1) * 100.0
data['log_ret'] = (np.log(data['Close'])).diff() * 100.0
data['gap'] = ((data['Open'] - data['Close'].shift()) / data['Open'] *
100.0)
data['gap2'] = ((data['Open'] - data['Close'].shift()) / data['Open'] *
100.0).shift()
data['gap3'] = ((data['Open'] - data['Close'].shift()) / data['Open'] *
100.0).shift(2)
data['gap4'] = ((data['Open'] - data['Close'].shift()) / data['Open'] *
100.0).shift(3)
data['gap5'] = ((data['Open'] - data['Close'].shift()) / data['Open'] *
100.0).shift(4)
data['hl'] = ((data['High'] - data['Low']) / data['Open'] * 100.0)
data['hl2'] = ((data['High'] - data['Low']) / data['Open'] * 100.0).shift()
data['hl3'] = ((data['High'] - data['Low']) / data['Open'] *
100.0).shift(2)
data['hl4'] = ((data['High'] - data['Low']) / data['Open'] *
100.0).shift(3)
data['hl5'] = ((data['High'] - data['Low']) / data['Open'] *
100.0).shift(4)
data['oc'] = ((data['Close'] - data['Open']) / data['Open'] * 100.0)
data['oc2'] = ((data['Close'] - data['Open']) / data['Open'] *
100.0).shift()
data['oc3'] = ((data['Close'] - data['Open']) / data['Open'] *
100.0).shift(2)
data['oc4'] = ((data['Close'] - data['Open']) / data['Open'] *
100.0).shift(3)
data['oc5'] = ((data['Close'] - data['Open']) / data['Open'] *
100.0).shift(4)
data['MA_short'] = talib.EMA(data['Close'].values, 10)
data['MA_long'] = talib.EMA(data['Close'].values, 120)
data['MA_ratio'] = (data['MA_short'] / data['MA_long'] - 1) * 100.0
data['MA2_short'] = talib.EMA(data['Close'].values, 10)
data['MA2_long'] = talib.EMA(data['Close'].values, 60)
data['MA2_ratio'] = (data['MA2_short'] / data['MA2_long'] - 1) * 100.0
data['vol_long'] = pd.rolling_std(data['Close'], 30)
data['vol_short'] = pd.rolling_std(data['Close'], 15)
data['vol_ratio'] = (data['vol_short'] / data['vol_long'] - 1) * 100.0
data['EMA'] = (Close / talib.EMA(Close, 5) - 1) * 100.0
data['EMA_long'] = (Close / talib.EMA(Close, 60) - 1) * 100.0
data['RSI'] = talib.RSI(data['Close'].values) / 100.0
data['MOM'] = talib.MOM(data['Close'].values, timeperiod=14) / 100.0
data['MACD_vfast'], data['MACD_signal_vfast'], data['MACD_hist'] = \
talib.MACD(data['Close'].values, fastperiod=4, slowperiod=9, signalperiod=3)
data['MACD_fast'], data['MACD_signal_fast'], _ = \
talib.MACD(data['Close'].values, fastperiod=12, slowperiod=26, signalperiod=9)
data['MACD_slow'], _, _ = talib.MACD(data['Close'].values,
fastperiod=25,
slowperiod=50)
data['MACD'], data['MACD_signal'], data['MACD_hist'] = talib.MACD(
data['Close'].values, fastperiod=30, slowperiod=65, signalperiod=22)
data['ATR'] = talib.ATR(High, Low, Close, timeperiod=28)
data['ADX_vlong'] = talib.ADX(High, Low, Close, timeperiod=120)
data['ADX_long'] = talib.ADX(High, Low, Close, timeperiod=28)
data['ADX_short'] = talib.ADX(High, Low, Close, timeperiod=14)
data['TSF_short'] = talib.TSF(data['Close'].values, timeperiod=14)
data['TSF_long'] = talib.TSF(data['Close'].values, timeperiod=28)
data['TSF_ratio'] = (data['TSF_short'] / data['TSF_long'] - 1) * 100.0
data['BBand_up'], data['BBand_mid'], data['BBand_low'] = talib.BBANDS(
data['Close'].values, timeperiod=20)
data['BBand_width'] = (data['BBand_up'] / data['BBand_low'] - 1) * 100.0
data['HMA_short'] = HMA(data['Close'].values, timeperiod=9)
data['HMA_long'] = HMA(data['Close'].values, timeperiod=60)
data['HMA_ratio'] = (data['HMA_short'] / data['HMA_long'] - 1) * 100.0
data['HMA_ret'] = HMA(data['Close'].values, 100)
# data['HMA_ret'] = data['HMA_ret'].pct_change()
data['OBV'] = talib.OBV(Close, Volume)
data['mean'] = pd.rolling_mean(data['ret'], 10)
data['std'] = pd.rolling_std(data['ret'], 10)
data['skewness'] = pd.rolling_skew(data['ret'], 10)
data['kurtosis'] = (pd.rolling_kurt(data['ret'], 10) - 3)
data['STOCHk'], data['STOCHd'] = talib.STOCH(High,
Low,
Close,
fastk_period=28,
slowk_period=3,
slowd_period=3)
data['STOCHRSId'], data['STOCHRSIk'] = talib.STOCHRSI(Close)
data['Chaikin_vol'] = Chaikin_vol(High, Low)
data['Chaikin_oscillator'] = Chaikin_oscillator(High, Low, Close, Volume)
data['PDI'] = talib.PLUS_DI(High, Low, Close, timeperiod=14)
data['MDI'] = talib.MINUS_DI(High, Low, Close, timeperiod=14)
data['DI'] = data['ADX_short'] - data['PDI'] + data['MDI']
# train_data = ['ret', 'ret_2', 'ret_3', 'ret_4', 'ret_5', 'vol_ratio', 'hl', 'oc', 'gap']
# 'ret_2', 'ret_3', 'ret_4', 'ret_5']
# data = include_VIX(data)
data.replace(np.nan, 0, inplace=True)
if normalization is True:
for feature in data.columns:
if feature not in [
'Open', 'High', 'Low', 'Close', 'Adj Close', 'Volume',
'Product', 'log_ret', 'ret', 'ret_2', 'ret_3', 'ret_4',
'ret_5', 'Date'
]:
data[feature] = (normalize(data[feature], start=start,
end=end))
if train_data is None:
# train_data = ['MACD_vfast', 'vol_ratio', 'oc', 'hl', 'ret', 'ADX_short', 'MA_ratio', 'MA2_ratio',
# 'RSI', 'skewness', 'kurtosis', 'mean', 'std']
train_data = ['oc', 'vol_ratio', 'hl', 'ret']
# train_data = ['MACD_vfast', 'vol_ratio', 'oc', 'hl', 'gap', 'ret', 'ADX_short', 'BBand_width', 'MA_ratio',
# 'RSI', 'skewness', 'kurtosis', 'mean', 'std'] # most original
# train_data = ['MACD_vfast', 'vol_ratio', 'oc', 'hl', 'gap', 'ret',
# 'ADX_short', 'BBand_width', 'MA_ratio', 'RSI', 'skewness', 'kurtosis', 'mean', 'std']
data = feature_analysis(data,
feature=train_data,
pca_components=len(train_data),
start=start,
end=end)
return data
def handle_data(context, data):
valid_contracts = [contract for contract in context.contracts
if contract.start_date <= get_datetime() - pd.Timedelta(days=SLOW_MA+2)
and contract.end_date >= get_datetime()]
hist = data.history(valid_contracts,
fields=['price', 'high', 'low'],
bar_count=SLOW_MA + 1,
frequency='1d')
slow_ma = hist['price'].apply(lambda x:
talib.EMA(x.as_matrix(),
timeperiod=SLOW_MA)[-1])
fast_ma = hist['price'].apply(lambda x:
talib.EMA(x.as_matrix(),
timeperiod=FAST_MA)[-1])
atr = hist.apply(lambda x: talib.ATR(x['high'].fillna(x['price']).as_matrix(),
x['low'].fillna(
x['price']).as_matrix(),
x['price'].as_matrix(),
timeperiod=SLOW_MA)[-1],
axis=(1, 0))
# breakout above is a buy signal
upper = hist['price'][-BREAKOUT-1:-2].max(axis=0)
# breakout below is a sell signal
lower = hist['price'][-BREAKOUT-1:-2].min(axis=0)
# last price
price = hist['price'].fillna(method='ffill').iloc[-1]
# position sizes as % of portfolio value
weights = RISK/100 * price/atr
longs = ((price > upper) & (fast_ma > slow_ma)) * weights
shorts = ((price < lower) & (fast_ma < slow_ma)) * -weights
signals = longs + shorts
# convert Continuous_future objects in indexes to current Future objects
signals.index = data.current(signals.index, 'contract')
atr.index = data.current(atr.index, 'contract')
price.index = data.current(price.index, 'contract')
signals = signals[signals != 0].dropna()
# rollover expiring contracts
for cont, pos in context.portfolio.positions.items():
if cont not in price.index:
weight = context.portfolio.current_portfolio_weights[cont]
root = cont.root_symbol
# close existing contract
signals[cont] = 0
current = data.current(continuous_future(root), 'contract')
position = context.portfolio.positions[cont]
# open current contract only if unrealised PnL positive
if (position.last_sale_price - position.cost_basis) * position.amount > 0:
if current: # contract could've been delisted
signals[current] = weight
# still need the old contract for stop-loss calculation
atr[cont] = atr[current]
price[cont] = price[current]
# implement stop-loss
for contract, position in context.portfolio.positions.items():
if contract in context.min_max:
context.min_max[contract] = (min(context.min_max[contract][0],
position.cost_basis,
position.last_sale_price),
max(context.min_max[contract][1],
position.cost_basis,
position.last_sale_price))
else:
context.min_max[contract] = (min(position.cost_basis,
position.last_sale_price),
max(position.cost_basis,
position.last_sale_price))
# calculate stop loss level
if position.amount > 0:
stop_price = context.min_max[contract][1] - \
atr[contract].item() * STOP
if price[contract] <= stop_price:
signals[contract] = 0
del context.min_max[contract]
if position.amount < 0:
stop_price = context.min_max[contract][0] + \
atr[contract].item() * STOP
if price[contract] >= stop_price:
signals[contract] = 0
del context.min_max[contract]
# execute trades
existing_positions = list(context.portfolio.positions.keys())
for asset, target in signals.items():
# pdb.set_trace()
log.info('transactions on {}:'.format(get_datetime()))
if asset in existing_positions:
# don't trade in existing positions unless it's stop loss
if target != 0:
continue
log.info('{}: {}'.format(asset, target))
order_target_percent(asset, target)
def on_message(client, userdata, msg):
logger.debug(msg.topic + " " + str(msg.payload))
global reading_watt_frigo, reading_watt_piastra, reading_watt_microwav, reading_watt_boiler
global time_watt_frigo, time_watt_piastra, time_watt_microwav, time_watt_boiler
global total_watt, total_watt_last, total_watt_var, power_watt_series, total_watt_last_ema
now = datetime.now()
value = json.loads(msg.payload)
watt = value["ENERGY"]["Power"]
t = datetime.strptime(value["Time"], TIME_FORMAT) + timedelta(
hours=TIMEZONE_IOT_OFFSET_HOURS)
if msg.topic == TOPIC_GR_FRIGO:
reading_watt_frigo = watt
time_watt_frigo = t
logger.debug("frigo W {:d} {}".format(reading_watt_frigo,
time_watt_frigo))
elif msg.topic == TOPIC_GR_PIASTRA:
reading_watt_piastra = watt
time_watt_piastra = t
logger.debug("piastr W {:d} {}".format(reading_watt_piastra,
time_watt_piastra))
elif msg.topic == TOPIC_GR_MICROWAV:
reading_watt_microwav = watt
time_watt_microwav = t
logger.debug("microw W {:d} {}".format(reading_watt_microwav,
time_watt_microwav))
elif msg.topic == TOPIC_GR_BOILER:
reading_watt_boiler = watt
time_watt_boiler = t
logger.debug("boiler W {:d} {}".format(reading_watt_boiler,
time_watt_boiler))
# scadenza valori
if (datetime.now() - time_watt_frigo).total_seconds() > EXPIRE_SEC:
time_watt_frigo = datetime.now()
reading_watt_frigo = 0
if (datetime.now() - time_watt_piastra).total_seconds() > EXPIRE_SEC:
time_watt_piastra = datetime.now()
reading_watt_piastra = 0
if (datetime.now() - time_watt_microwav).total_seconds() > EXPIRE_SEC:
time_watt_microwav = datetime.now()
reading_watt_microwav = 0
if (datetime.now() - time_watt_boiler).total_seconds() > EXPIRE_SEC:
time_watt_boiler = datetime.now()
reading_watt_boiler = 0
# calcolo
total_watt = reading_watt_frigo + reading_watt_piastra + reading_watt_microwav + reading_watt_boiler
total_watt_var = round(
-(total_watt_last - total_watt) / TOTAL_WATT_MAX * 100.0, 1)
total_watt_perc = round(total_watt / TOTAL_WATT_MAX * 100.0, 1)
power_watt_series = np.delete(power_watt_series, 0)
power_watt_series = np.append(power_watt_series, total_watt)
# EMA = exponential moving average
# FIR filter
total_watt_ema = round(talib.EMA(power_watt_series, timeperiod=4)[10 - 1])
total_watt_var_ema = round(
-(total_watt_last_ema - total_watt_ema) / TOTAL_WATT_MAX * 100.0, 1)
# if total_watt_var > 200:
logger.info(
"total W {:d} var % {} load % {} EMA4 W {} var EMA4 % {}".format(
total_watt, total_watt_var, total_watt_perc, total_watt_ema,
total_watt_var_ema))
# update
total_watt_last = total_watt
total_watt_last_ema = total_watt_ema
msg = None
if total_watt_var_ema > MESSAGE_EMA_LEVEL:
msg = "la potenza sta salendo al {} percento".format(
round(total_watt_perc))
if total_watt_var_ema < -MESSAGE_EMA_LEVEL:
msg = "la potenza sta scendendo al {} percento".format(
round(total_watt_perc))
if msg is not None:
# tts @ home assistant
payload = {
"time": now.strftime(TIME_FORMAT),
"speech": {
"entity_id": HA_CHROMECAST_ID,
"language": 'it',
"message": msg
}
}
client.publish(HA_SERVICE, json.dumps(payload))
timeperiod = 120 # 하이퍼파라미터 몇가지를 잘 모르겠다. # Turnover # Stock Turnover = Cost of Goods Sold / Average Inventory, # 거래량 / 총 발행 주식수 # Bias ??? # Bollingerbands df_amd_bb_upper, df_amd_bb_middle, df_amd_bb_lower = ta.BBANDS(df_amd['Close'], timeperiod=timeperiod) # Directionalmovementindex df_amd_dx = ta.DX(df_amd['Close'],df_amd['Open'],df_amd['High'], timeperiod=timeperiod) # Exponentialmovingaverages df_amd_ema = ta.EMA(df_amd['Close'], timeperiod=timeperiod) # Stochasticindex df_amd_slowk, df_amd_slowd = ta.STOCH(df_amd['High'], df_amd['Low'], df_amd['Close'], fastk_period=5, slowk_period=3, slowk_matype=0, slowd_period=3, slowd_matype=0) # Movingaverages df_amd_ma5 = ta.MA(df_amd['Close'], timeperiod=timeperiod) # MACD df_amd_macd, df_amd_macdsignal, df_amd_macdhist = ta.MACD(df_amd['Close'], fastperiod=12, slowperiod=26, signalperiod=9) # Relativestrengthindex df_amd_rsi = ta.RSI(df_amd['Close'], timeperiod=timeperiod) df_amd_total = df_amd.copy()
def get_dataframe(index_component, input_window_length, forecast_horizon,
start_date, end_date, frequency, classnum, method):
#get dataframe from sigle company
#input shortcuts
iwl = input_window_length
fh = forecast_horizon
method_list = ['ANN', 'SVM']
if not (method in method_list):
raise Exception(method + ' is not in' + str(method_list))
ic = index_component
this_ti = get_price(ic, start_date, end_date, frequency=frequency)
dates = this_ti.index[:-fh]
close = np.array(this_ti['close'])
next_close = get_next_close(close, fh)
high = np.array(this_ti['high'])
low = np.array(this_ti['low'])
Y = labeling(fh, close[:-fh], next_close, classes=classnum)
sma = talib.SMA(close, iwl)[:-fh]
ema = talib.EMA(close, iwl)[:-fh]
atr = talib.ATR(high, low, close, iwl)[:-fh]
admi = talib.ADX(high, low, close, iwl)[:-fh]
cci = talib.CCI(high, low, close, iwl)[:-fh]
roc = talib.ROC(close, timeperiod=iwl)[:-fh]
rsi = talib.RSI(close, timeperiod=iwl)[:-fh]
williams = talib.WILLR(high, low, close, timeperiod=iwl)[:-fh]
slowk = stochasticK(high, low, close, iwl)[:-fh]
slowd = talib.EMA(slowk, 3)
#form a initial data frame
df = pd.DataFrame(
{
'close price t': close[:-fh],
'close price t+s': next_close,
# 'Y':Y,
'SMA': sma,
'EMA': ema,
'ATR': atr,
'ADMI': admi,
'CCI': cci,
'ROC': roc,
'RSI': rsi,
'Williams %R': williams,
'Stochastic %K': slowk,
'Stochastic %D': slowd,
},
index=dates)
# append label according to the method
if method == 'SVM':
df['Y'] = Y
df = df.dropna(axis=0, how='any')
y = df['Y'].values
if method == 'ANN':
one_hot_code = get_onehotcode(Y)
ys = []
for i in range(len(one_hot_code[0])):
string = 'Y' + str(i + 1)
ys.append(string)
df[string] = one_hot_code[:, i]
df = df.dropna(axis=0, how='any')
y = df[ys].values
x = df[[
'ADMI', 'ATR', 'CCI', 'EMA', 'ROC', 'RSI', 'SMA', 'Stochastic %D',
'Stochastic %K', 'Williams %R'
]].values
return x, y
def test_ema():
c = numpy.random.randn(100)
# this is the library function
d = talib.EMA(c, timeperiod=30)
assert d is not None
def EMA(df):
df['EMA7'] = talib.EMA(df.close, timeperiod=5)
df['EMA14'] = talib.EMA(df.close, timeperiod=10)
return df
def EMA(self, window=30):
real = talib.EMA(self.close, timeperiod=window)
return real
def all_technical_indicator_inputs(
ohlc_data: pd.DataFrame) -> pd.DataFrame:
inputs = {
'open': ohlc_data['Open'],
'high': ohlc_data['High'],
'low': ohlc_data['Low'],
'close': ohlc_data['Close'],
'volume': ohlc_data['Volume']
}
# Price Data
def day_diff(diff_days: int):
n_day_diff = ohlc_data["Close"][diff_days:].to_numpy(
) - ohlc_data["Close"][:-diff_days].to_numpy()
n_day_diff = np.insert(n_day_diff, 0, [np.nan] * diff_days, axis=0)
return n_day_diff
for i in range(5):
k = i + 1
ohlc_data["diff-day-" + str(k)] = day_diff(k)
# Simple Indicators
ohlc_data["EMA-2"] = ohlc_data["Close"] - talib.EMA(
ohlc_data["Close"], 2)
ohlc_data["EMA-3"] = ohlc_data["Close"] - talib.EMA(
ohlc_data["Close"], 3)
ohlc_data["EMA-5"] = ohlc_data["Close"] - talib.EMA(
ohlc_data["Close"], 5)
ohlc_data["EMA-7"] = ohlc_data["Close"] - talib.EMA(
ohlc_data["Close"], 7)
ohlc_data["EMA-21"] = ohlc_data["Close"] - talib.EMA(
ohlc_data["Close"], 21)
ohlc_data["EMA-50"] = ohlc_data["Close"] - talib.EMA(
ohlc_data["Close"], 50)
ohlc_data["RSI"] = talib.RSI(ohlc_data["Close"], 14) / 100
ohlc_data["MACD"] = talib.MACD(ohlc_data["Close"], 12, 26, 9)[2]
ohlc_data["ADX"] = talib.ADX(ohlc_data["High"], ohlc_data["Low"],
ohlc_data["Close"]) / 100
ohlc_data["STD"] = talib.STDDEV(ohlc_data["Close"])
# Raw Indicators
ohlc_data["RAW-EMA-2"] = talib.EMA(ohlc_data["Close"], 2)
ohlc_data["RAW-EMA-3"] = talib.EMA(ohlc_data["Close"], 3)
ohlc_data["RAW-EMA-5"] = talib.EMA(ohlc_data["Close"], 5)
ohlc_data["RAW-EMA-7"] = talib.EMA(ohlc_data["Close"], 7)
ohlc_data["RAW-EMA-21"] = talib.EMA(ohlc_data["Close"], 21)
ohlc_data["RAW-EMA-50"] = talib.EMA(ohlc_data["Close"], 50)
# Complex Indicators
def get_useful_bbands():
bb = talib.BBANDS(ohlc_data["Close"])
bb_width = bb[0] - bb[2]
bb_buy = ohlc_data["Close"] - bb[0]
bb_sell = bb[2] - ohlc_data["Close"]
return bb_width, bb_buy, bb_sell
ohlc_data["bb_width"], ohlc_data["bb_buy"], ohlc_data[
"bb_sell"] = get_useful_bbands()
ohlc_data['Volume'] = ohlc_data['Volume'].diff()
return ohlc_data
def cal_ema(close, peroid):
return talib.EMA(np.array(close), timeperiod=peroid)
def handle_data(context, data):
short_window = 5 #value obtained from Belinda script
long_window = 24 #value obtained from Belinda script
context.i += 1
if context.i < long_window:
return
#calculate EMA with Talib
my_short_stock_series = data.history(context.asset, 'price', bar_count = short_window, frequency = "1T",)
ema_short_result = talib.EMA(my_short_stock_series, timeperiod=short_window)
short_ema = ema_short_result[-1]
my_long_stock_series = data.history(context.asset, 'price', bar_count = long_window, frequency = "1T",)
ema_long_result = talib.EMA(my_long_stock_series, timeperiod=long_window)
long_ema = ema_long_result[-1]
price = data.current(context.asset, 'price')
if context.base_price is None:
context.base_price = price
price_change = (price - context.base_price)/context.base_price
record(price = price, cash = context.portfolio.cash, price_change = price_change, short_ema = short_ema, long_ema = long_ema)
pos_amount = context.portfolio.positions[context.asset].amount
orders = context.blotter.open_orders
if len(orders) > 0:
return
if not data.can_trade(context.asset):
return
#sizing calculations
#getting high, low, close value from history to talib.ATR function
highs = data.history(context.asset, 'high', bar_count = 300, frequency = "1T",).dropna()
lows = data.history(context.asset, 'low', bar_count = 300, frequency = "1T",).dropna()
closes = data.history(context.asset, 'close', bar_count = 300, frequency = "1T",).dropna()
atr = talib.ATR(highs, lows, closes) #talib.ATR output is an array
stop_loss_atr = atr[-1] #talib.ATR output is an array
stop = stop_loss_atr
#print("stop_loss_atr=", stop_loss_atr)
#position size calculation based on current equity, risk and trailing stop
size = context.portfolio.portfolio_value*context.risk/stop
trading_avaliable = context.portfolio.portfolio_value/0.3/price
size_aval = size if size<trading_avaliable else trading_avaliable #equal to PineScript statement, size_aval = size<trading_avaliable?size:trading_avaliable
#print("size_aval=", size_aval)
#trailing stop call. obtaining context.stop_price
set_trailing_stop(context, data, stop) #maybe trailing stop should be calculated when order start
set_trailing_stop_short(context, data, stop)
#print("price=", price, "context.stop_price=", context.stop_price,"context.stop_price_short=", context.stop_price_short)
#check for the long stop price and long orders are open
if (pos_amount > 0):
if price < context.stop_price:
order_target_percent(context.asset, 0)
context.stop_price = 0
#print("trailing stop")
if (pos_amount < 0):
if price > context.stop_price_short:
order_target_percent(context.asset, 0)
context.stop_price_short = 0
#print("trailing stop short")
if short_ema > long_ema and pos_amount <=0:
order_target_percent(context.asset, 0) #close short asset
order(context.asset, size_aval) #open long asset
#print("long")
elif short_ema < long_ema and pos_amount > 0:
order_target_percent(context.asset, 0) #close long asset
order(context.asset, -size_aval) #open short asset
def ema(self, n, array=False):
"""简单均线"""
result = talib.EMA(self.close, n)
if array:
return result
return result[-1]
output = pd.concat([leo_candles, leodf_resampled], axis=1)
output['burn_amount_usd'] = output['leo_burn_amount'] * output['leo_open']
btc_candles = pd.read_csv('data/BTCUSD_3H_2018-01-01-present.csv', parse_dates=[0], infer_datetime_format=True)
btc_candles.set_index(btc_candles['timestamp'], inplace=True)
btc_candles.sort_index(inplace=True)
btc_candles.rename(columns={'open': 'btc_open', 'volume': 'btc_volume'}, inplace=True)
btc_candles.drop(['timestamp','high','low','close'], axis=1, inplace=True)
btc_candles['btc_usd_volume'] = btc_candles['btc_volume'] * btc_candles['btc_open']
output = pd.concat([output, btc_candles], axis=1)
output['burn_amount_usd_EMA'] = ta.EMA(output['burn_amount_usd'].ffill(), ema_period)
output['btc_usd_volume_EMA'] = ta.EMA(output['btc_usd_volume'], ema_period)
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import matplotlib.dates as mdates
from matplotlib.lines import Line2D
from matplotlib.patches import Rectangle, Patch
import matplotlib.patches as mpatches
# from matplotlib.finance import candlestick_ohlc
from matplotlib.ticker import ScalarFormatter, NullFormatter
'/Users/shashank/Documents/Code/Python/Finance/spxTickers.pickle')
tickers = [item.replace(".", "-") for item in tickers]
frames = []
for symbol in tickers:
ticker = ['{}'.format(symbol)]
# Read data
data = yf.download(symbol, start, end)
# ## SMA and EMA
#Simple Moving Average
data['SMA'] = talib.SMA(data['Adj Close'], timeperiod=20)
# Exponential Moving Average
data['EMA'] = talib.EMA(data['Adj Close'], timeperiod=20)
# Bollinger Bands
data['upper_band'], data['middle_band'], data['lower_band'] = talib.BBANDS(
data['Adj Close'], timeperiod=20)
# ## MACD (Moving Average Convergence Divergence)
# MACD
data['macd'], data['macdsignal'], data['macdhist'] = talib.MACD(
data['Adj Close'], fastperiod=12, slowperiod=26, signalperiod=9)
# ## RSI (Relative Strength Index)
# RSI
data['RSI'] = talib.RSI(data['Adj Close'], timeperiod=14)
# ## OBV (On Balance Volume)
def run():
"""
* 如果趋势发生变化，撤销所有的限价挂单以及委托挂单
* 如果有逆势持仓，则平掉所有逆势持仓
* 如果有逆势限价挂单，则撤销所有逆势限价挂单
* 如果没有顺势持仓，则（如果没有限价挂单，则按最优价格挂单；如果有限价挂单，检查价位是否合理，如果不合理则撤销重新挂单）
* 如果有顺势持仓，则只挂止盈委托挂单，在慢速ema价格上设置止盈点
"""
global g_ret
g_ret = qds_test_registration()
if not g_ret:
logging.debug("Error, software is not correctly registered. Please contact [email protected] to register")
return False
global period
global ma_fast
global ma_slow
global open_offset
global open_interval
global stop_offset
global level_rate
global max_open_number
global trend_history
global_data = Organized()
logging.debug("qds params: period={0}, ma_fast={1}, ma_slow={2}, open_offset={3}, open_interval={4}, "
"stop_offset={5}, level_rate={6}, max_open_number={7}".format(
period, ma_fast, ma_slow, open_offset, open_interval, stop_offset, level_rate, max_open_number))
if not get_system_running(): return False
g_ret = dm.get_contract_kline(symbol=symbol_period, period=period, size=300)
if not ru.is_ok(g_ret):
logging.debug("get_contract_kline failed, ret={0}".format(pformat(g_ret)))
return False
global_data = KLineAdapterHuobi.ParseData(g_ret['data'])
close_list = global_data._close_list
ma_fast_list = talib.EMA(np.array(close_list), timeperiod=ma_fast)
ma_slow_list = talib.EMA(np.array(close_list), timeperiod=ma_slow)
global_data._ema_list[ma_fast] = ma_fast_list
global_data._ema_list[ma_slow] = ma_slow_list
# debug_ema(global_data)
last_index = global_data.GetLen() - 2
last_ema_fast = global_data.ema_list[ma_fast][last_index]
last_ema_slow = global_data.ema_list[ma_slow][last_index]
ts = global_data._timestamp[last_index]
# 多空决策
trend = None
if last_ema_fast > last_ema_slow:
trend = 'long'
elif last_ema_fast < last_ema_slow:
trend = 'short'
else: # 趋势不变
if trend_history:
trend = trend_history
else:
trend = 'long'
# 趋势发生变化，撤销所有的限价挂单以及委托挂单
logging.debug("ts:{0} ma{1}:{2}, ma{3}:{4}".format(ts, ma_fast, last_ema_fast, ma_slow, last_ema_slow))
if trend_history != trend:
logging.debug(
"trend={5} last: {4} ma{0}:{1}, ma{2}:{3}".format(ma_fast, last_ema_fast, ma_slow, last_ema_slow, ts,
trend))
if not get_system_running(): return False
g_ret = dm.cancel_all_contract_order(symbol_type)
if ru.is_ok(g_ret):
logging.debug("cancel_all_contract_order successfully at trend changed")
else:
if ru.is_no_order(g_ret):
logging.debug("cancel_all_contract_order no orders")
else:
logging.debug("cancel_all_contract_order failed at trend changed, ret={0}".format(pformat(g_ret)))
return False
if not get_system_running(): return False
g_ret = dm.cancel_all_contract_trigger(symbol_type)
if ru.is_ok(g_ret):
logging.debug("cancel_all_contract_trigger successfully at trend changed")
else:
if ru.is_no_order(g_ret):
logging.debug("cancel_all_contract_trigger no orders")
else:
logging.debug("cancel_all_contract_trigger failed at trend changed, ret={0}".format(pformat(g_ret)))
return False
# 确保撤单成功再修改趋势，否则下次会再次执行撤单操作
trend_history = trend
if not get_system_running(): return False
g_ret = dm.get_contract_account_position_info(symbol_type)
if ru.is_ok(g_ret):
available = g_ret['data'][0]['margin_available']
balance = g_ret['data'][0]['margin_balance']
logging.debug("margin_available={0}, margin_balance={1}".format(available, balance))
set_margin(available, balance)
if available == 0.0 and balance == 0.0:
logging.debug("no available margin {0}, {1}".format(available, balance))
return False
# 获取当前持仓多单数量，空单数量，价格
if not get_system_running(): return False
g_ret = dm.get_contract_position_info(symbol_type)
if not ru.is_ok(g_ret):
logging.debug("get_contract_position_info failed, ret={0}".format(pformat(g_ret)))
return False
cpi_helper.log_all_orders("buy", g_ret)
cpi_helper.log_all_orders('sell', g_ret)
limit_holding_buy_count = cpi_helper.get_orders_count('buy', g_ret)
limit_holding_buy_price = cpi_helper.get_price('buy', g_ret)
limit_holding_sell_count = cpi_helper.get_orders_count('sell', g_ret)
limit_holding_sell_price = cpi_helper.get_price('sell', g_ret)
# 获取当前限价挂单的方向以及价格
if not get_system_running(): return False
g_ret = dm.get_contract_open_orders(symbol_type)
if not ru.is_ok(g_ret):
logging.debug("get_contract_open_orders failed, ret={0}".format(pformat(g_ret)))
return False
coo_helper.log_all_orders('buy', g_ret)
coo_helper.log_all_orders('sell', g_ret)
limit_pending_buy_count = coo_helper.get_orders_count('buy', 'open', g_ret) # 开仓
limit_pending_close_sell_count = coo_helper.get_orders_count('sell', 'close', g_ret) # 平仓
limit_pending_sell_count = coo_helper.get_orders_count('sell', 'open', g_ret)
limit_pending_close_buy_count = coo_helper.get_orders_count('buy', 'close', g_ret)
limit_pending_buy_price_list = coo_helper.get_price('buy', 'open', g_ret)
limit_pending_sell_price_list = coo_helper.get_price('sell', 'open', g_ret)
# 获取当前委托挂单多单数量，空单数量
if not get_system_running(): return False
g_ret = dm.get_contract_trigger_openorders(symbol_type)
if not ru.is_ok(g_ret):
logging.debug("get_contract_trigger_openorders failed, ret={0}".format(pformat(g_ret)))
return False
ctoo_helper.log_all_orders('buy', g_ret)
ctoo_helper.log_all_orders('sell', g_ret)
trigger_holding_buy_count = ctoo_helper.get_orders_count('buy', 'close', g_ret)
trigger_holding_sell_count = ctoo_helper.get_orders_count('sell', 'close', g_ret)
# 获取当前委托挂单方向以及价格
trigger_holding_buy_order_price = round(ctoo_helper.get_order_price('buy', g_ret))
trigger_holding_sell_order_price = round(ctoo_helper.get_order_price('sell', g_ret))
# 设置最大允许操作数量(挂单数量+持仓数量)
max_on_trend_count = max_open_number
# 计算当前是否有逆势持仓
limit_holding_against_trend_count = 0
limit_holding_against_trend_close_direction = ''
limit_holding_price = 0
trigger_holding_against_trend_count = 0
# 止盈止损
stop_earning_trigger_type = ''
stop_earning_trigger_price = 0
stop_earning_order_price = 0
stop_earning_direction = ''
limit_pending_on_trend_count = 0
limit_pending_close_on_trend_count = 0
limit_pending_against_trend_count = 0
limit_pending_on_trend_price_list = []
trigger_holding_on_trend_order_price = 0
base_open_point = 0
if trend == 'long':
base_open_point = last_ema_slow - open_offset
limit_holding_against_trend_count = limit_holding_sell_count
limit_holding_against_trend_close_direction = 'buy'
trigger_holding_against_trend_count = trigger_holding_sell_count
trigger_holding_on_trend_count = trigger_holding_sell_count
trigger_holding_on_trend_order_price = trigger_holding_sell_order_price
limit_pending_on_trend_open_direction = 'buy'
limit_pending_on_trend_count = limit_pending_buy_count
limit_pending_close_on_trend_count = limit_pending_close_sell_count
limit_pending_against_trend_count = limit_pending_sell_count
limit_pending_on_trend_price_list = limit_pending_buy_price_list
limit_holding_on_trend_count = limit_holding_buy_count
limit_holding_price = limit_holding_buy_price
stop_earning_trigger_type = 'ge'
stop_earning_order_price = round(last_ema_slow + stop_offset)
stop_earning_trigger_price = round(stop_earning_order_price - 10)
stop_earning_direction = 'sell'
else:
base_open_point = last_ema_slow + open_offset
limit_holding_against_trend_count = limit_holding_buy_count
limit_holding_against_trend_close_direction = 'sell'
trigger_holding_against_trend_count = trigger_holding_buy_count
trigger_holding_on_trend_count = trigger_holding_buy_count
trigger_holding_on_trend_order_price = trigger_holding_buy_order_price
limit_pending_on_trend_open_direction = 'sell'
limit_pending_on_trend_count = limit_pending_sell_count
limit_pending_close_on_trend_count = limit_pending_close_buy_count
limit_pending_against_trend_count = limit_pending_buy_count
limit_pending_on_trend_price_list = limit_pending_sell_price_list
limit_holding_on_trend_count = limit_holding_sell_count
limit_holding_price = limit_holding_sell_price
stop_earning_trigger_type = 'le'
stop_earning_order_price = round(last_ema_slow - stop_offset)
stop_earning_trigger_price = round(stop_earning_order_price + 10)
stop_earning_direction = 'buy'
# 最优限价挂单价格
limit_best_price = [
round(base_open_point - open_interval),
round(base_open_point),
round(base_open_point + open_interval)
]
logging.debug("limit_best_price={0}, {1}, {2}".format(limit_best_price[0], limit_best_price[1], limit_best_price[2]))
# 执行交易
# 平掉所有逆势持仓
if limit_holding_against_trend_count > 0:
logging.debug("limit_holding_against_trend_count={0}".format(limit_holding_against_trend_count))
if not get_system_running(): return False
g_ret = dm.send_lightning_close_position(symbol_type, contract_type_period, '',
int(limit_holding_against_trend_count),
limit_holding_against_trend_close_direction,
'', None)
if ru.is_ok(g_ret):
limit_holding_against_trend_count = 0
logging.debug("send_lightning_close_position successfully, direction={0}, volume={1}".format(
limit_holding_against_trend_close_direction, limit_holding_against_trend_count))
else:
logging.debug("send_lightning_close_position failed, ret={0}".format(pformat(g_ret)))
return False
# 撤销所有逆势限价挂单，简化操作，只要逆势挂单就撤销所有限价挂单
if limit_pending_against_trend_count > 0:
logging.debug("limit_pending_against_trend_count={0}".format(limit_pending_against_trend_count))
if not get_system_running(): return False
g_ret = dm.cancel_all_contract_order(symbol_type)
if ru.is_ok(g_ret):
limit_pending_on_trend_count = 0
limit_pending_against_trend_count = 0
logging.debug("cancel_all_contract_order successfully")
else:
logging.debug("cancel_all_contract_order failed, ret={0}".format(pformat(g_ret)))
return False
# 如果没有顺势持仓，则（如果没有限价挂单，则按最优价格挂单；如果有限价挂单，检查价位是否合理，如果不合理则撤销重新挂单）
if limit_holding_on_trend_count == 0:
if limit_pending_on_trend_count == 0:
available_limit_pending_on_trend_count = max_on_trend_count
logging.debug("available_limit_pending_on_trend_count={0}".format(available_limit_pending_on_trend_count))
if available_limit_pending_on_trend_count > 0:
available_limit_pending_on_trend_count_each = [
round(available_limit_pending_on_trend_count / 3),
round(available_limit_pending_on_trend_count / 3),
max_on_trend_count - round(available_limit_pending_on_trend_count / 3) * 2
]
for i in range(0, len(available_limit_pending_on_trend_count_each)):
if limit_holding_on_trend_count < max_on_trend_count:
price = round(limit_best_price[i])
volume = available_limit_pending_on_trend_count_each[i]
if volume > available_limit_pending_on_trend_count:
volume = available_limit_pending_on_trend_count
direction = limit_pending_on_trend_open_direction
if price > 0 and volume > 0:
if not get_system_running(): return False
g_ret = dm.send_contract_order(symbol=symbol_type, contract_type=contract_type_period, contract_code='',
client_order_id='', price=price, volume=int(volume),
direction=direction, offset='open', lever_rate=level_rate,
order_price_type='limit')
if ru.is_ok(g_ret):
available_limit_pending_on_trend_count -= volume
logging.debug(
"send_contract_order successfully, price={0} volume={1} direction={2}".format(
price, int(volume), direction))
else:
logging.debug("send_contract_order failed, ret={0}".format(pformat(g_ret)))
return False
elif limit_pending_on_trend_count != max_on_trend_count:
global g_fix_race_check_one_more_time
if limit_pending_on_trend_count < max_on_trend_count and not g_fix_race_check_one_more_time:
logging.debug(
"Fix race issue: limit_pending_on_trend_count {0} is smaller than max_on_trend_count {1}, "
"need to check one more time".format(limit_pending_on_trend_count, max_on_trend_count))
g_fix_race_check_one_more_time = True
else:
g_fix_race_check_one_more_time = False
if not get_system_running(): return False
g_ret = dm.cancel_all_contract_order(symbol_type)
if ru.is_ok(g_ret):
logging.debug(
"cancel_all_contract_order successfully, limit_pending_on_trend_count {0} not match with "
"max_on_trend_count {1}".format(limit_pending_on_trend_count, max_on_trend_count))
else:
logging.debug("cancel_all_contract_order failed, ret={0}".format(pformat(g_ret)))
return False
else:
logging.debug("limit_pending_on_trend_count={0}".format(limit_pending_on_trend_count))
for i in range(0, len(limit_pending_on_trend_price_list)):
price = limit_pending_on_trend_price_list[i]
# 如果限价挂单价格不是最优价格，则撤销重新挂单
is_best_limit_price = False
for j in range(0, len(limit_best_price)):
if price == limit_best_price[j]:
is_best_limit_price = True
if not is_best_limit_price:
if not get_system_running(): return False
g_ret = dm.cancel_all_contract_order(symbol_type)
if ru.is_ok(g_ret):
logging.debug(
"cancel_all_contract_order successfully, price={0} is not one of limit_best_price".format(
price))
else:
logging.debug("cancel_all_contract_order failed, ret={0}".format(pformat(g_ret)))
return False
break
# 如果有顺势持仓，则只挂止盈委托挂单，在持仓的价格上设置止盈点
# 在只有限价单持仓的情况下，才能设置止盈委托挂单，否则会容易交易失败，而且不合逻辑
available_trigger_pending_on_trend_count = limit_holding_on_trend_count - \
limit_pending_close_on_trend_count - trigger_holding_on_trend_count
if available_trigger_pending_on_trend_count > 0:
logging.debug("available_trigger_pending_on_trend_count={0}".format(available_trigger_pending_on_trend_count))
if not get_system_running(): return False
g_ret = dm.send_contract_trigger_order(symbol=symbol_type, contract_type=contract_type_period,
contract_code=None,
trigger_type=stop_earning_trigger_type,
trigger_price=round(stop_earning_trigger_price),
order_price=round(stop_earning_order_price),
order_price_type='limit',
volume=int(available_trigger_pending_on_trend_count),
direction=stop_earning_direction,
offset='close',
lever_rate=level_rate)
if ru.is_ok(g_ret):
logging.debug("send_contract_trigger_order successfully, {0} {1} {2} {3} {4}".format(
stop_earning_trigger_type, round(stop_earning_trigger_price), round(stop_earning_order_price),
int(available_trigger_pending_on_trend_count), stop_earning_direction))
else:
logging.debug("send_contract_trigger_order failed, ret={0}".format(pformat(g_ret)))
return False
def handle_data(context):
# 买入信号
long_signal_triggered = False
# 卖出信号
short_signal_triggered = False
# 止损信号
stop_loss_signal_triggered = False
# 更新净值最大值
if context.user_data.max_net is None:
context.user_data.max_net = context.account.huobi_cny_net
else:
if context.user_data.max_net < context.account.huobi_cny_net:
context.user_data.max_net = context.account.huobi_cny_net
# 计算当前回撤
current_draw_down = (
1 - context.account.huobi_cny_net / context.user_data.max_net) * 100
context.log.info("当前回撤为 %.2f%%, 止损线为 %.2f%%" %
(current_draw_down, context.user_data.stop_loss_line))
# 当前回撤大于止损线，则产生卖出止损信号
if current_draw_down > context.user_data.stop_loss_line:
context.log.info("已经触发止损线，全仓卖出止损，等待下一次买入信号")
stop_loss_signal_triggered = True
# 获取历史数据
hist = context.data.get_price(context.security,
count=context.user_data.ema_slow_window + 1,
frequency=context.frequency)
if len(hist.index) < context.user_data.ema_slow_window:
context.log.warn("bar的数量不足, 等待下一根bar...")
return
# 收盘价
close_prices = hist["close"].values
# 计算EMA值
ema_fast = talib.EMA(close_prices, context.user_data.ema_fast_window)
ema_slow = talib.EMA(close_prices, context.user_data.ema_slow_window)
# 当前快线EMA
current_ema_fast = ema_fast[-1]
# 当前慢线EMA
current_ema_slow = ema_slow[-1]
# 前一个bar的快线EMA
pre_ema_fast = ema_fast[-2]
# 前一个bar的慢线EMA
pre_ema_slow = ema_slow[-2]
context.log.info(
"当前EMA 快线 = %.2f, 慢线 = %.2f; 前一个bar EMA 快线 = %.2f, 慢线 = %.2f" %
(current_ema_fast, current_ema_slow, pre_ema_fast, pre_ema_slow))
# EMA快线从下向上穿过EMA慢线时，产生买入信号
if pre_ema_fast <= pre_ema_slow and current_ema_fast > current_ema_slow:
context.log.info("EMA快线从下向上穿过EMA慢线时，产生买入信号")
long_signal_triggered = True
# EMA快线从上向下穿过EMA慢线时，产生卖出信号
elif pre_ema_fast >= pre_ema_slow and current_ema_fast < current_ema_slow:
context.log.info("EMA快线从上向下穿过EMA慢线时，产生卖出信号")
short_signal_triggered = True
# 有卖出信号，且持有仓位，则市价单全仓卖出
if short_signal_triggered or stop_loss_signal_triggered:
if context.account.huobi_cny_btc >= HUOBI_CNY_BTC_MIN_ORDER_QUANTITY:
context.user_data.max_net = None
# 卖出信号，且不是空仓，则市价单全仓清空
context.log.info("正在卖出 %s" % context.security)
context.log.info("卖出数量为 %s" % context.account.huobi_cny_btc)
context.order.sell(context.security,
quantity=str(context.account.huobi_cny_btc))
else:
context.log.info("仓位不足，无法卖出")
# 有买入信号，且持有现金，则市价单全仓买入
elif long_signal_triggered:
if context.account.huobi_cny_cash >= HUOBI_CNY_BTC_MIN_ORDER_CASH_AMOUNT:
# 买入信号，且持有现金，则市价单全仓买入
context.log.info("正在买入 %s" % context.security)
context.log.info("下单金额为 %s 元" % context.account.huobi_cny_cash)
context.order.buy(context.security,
cash_amount=str(context.account.huobi_cny_cash))
else:
context.log.info("现金不足，无法下单")
else:
context.log.info("无交易信号，进入下一根bar")
def Chaikin_oscillator(High, Low, Close, Volume, n=3, m=9):
ADL = ((Close - Low) - (High - Close)) / (High - Low) * Volume + 1
return talib.EMA(ADL, n) - talib.EMA(ADL, m)
