def label_func(data, snp_vector, vector, complete_datum, index, window=14):
if len(snp_vector) <= index + window:
return None
windowed_index = index + window
roc_vector = tl.ROC(vector, timeperiod=window)
snp_roc = tl.ROC(snp_vector, timeperiod=window)
beta = tl.BETA(roc_vector, snp_roc, timeperiod=window)
expected_return = beta[windowed_index] * (snp_roc[windowed_index] - RISK_FREE)
excess = beta[windowed_index] * (snp_roc[windowed_index] - RISK_FREE) + RISK_FREE
label = 'buy' if data[windowed_index]['Open'] < data[windowed_index]['Close'] else 'stay'
#print 'stock roc: %f, SnP roc: %f, beta: %f, exp-ret.: %f' %\
#(roc_vector[index], snp_roc[index], beta[index], RISK_FREE + beta[index] * (snp_roc[index] - RISK_FREE)))
if expected_return > 0:
bx = 1
else:
bx = -1
complete_datum['bx'] = bx
complete_datum['excess'] = excess
complete_datum['label'] = label
return complete_datum
def process_raw(self, csv_file):
df = self._file
df['dt'] = df['Date'].astype(str) + " " + df['Time'].astype(str)
df['DateTime'] = pd.to_datetime(df['dt'],
format=' % Y % m % d % H: % M: % S',
errors='ignore')
df = df.set_index(['DateTime'])
df = df.drop(['General', 'Date', 'Time', 'dt'], axis=1)
df['EMA10'] = ta.EMA(df['Close'], timeperiod=10)
df['EMA30'] = ta.EMA(df['Close'], timeperiod=30)
df['EMA60'] = ta.EMA(df['Close'], timeperiod=60)
df['SMA10'] = ta.SMA(df['Close'], timeperiod=10)
df['SMA30'] = ta.SMA(df['Close'], timeperiod=30)
df['SMA60'] = ta.SMA(df['Close'], timeperiod=60)
df['WMA10'] = ta.WMA(df['Close'], timeperiod=10)
df['WMA30'] = ta.WMA(df['Close'], timeperiod=30)
df['WMA60'] = ta.WMA(df['Close'], timeperiod=60)
df['RSI10'] = ta.RSI(df['Close'], timeperiod=10)
df['RSI30'] = ta.RSI(df['Close'], timeperiod=30)
df['RSI60'] = ta.RSI(df['Close'], timeperiod=60)
df['MOM10'] = ta.MOM(df['Close'], timeperiod=10)
df['MOM30'] = ta.MOM(df['Close'], timeperiod=30)
df['MOM60'] = ta.MOM(df['Close'], timeperiod=60)
df['ROC10'] = ta.ROC(df['Close'], timeperiod=10)
df['ROC50'] = ta.ROC(df['Close'], timeperiod=30)
df['ROC60'] = ta.ROC(df['Close'], timeperiod=60)
df.to_csv(csv_file)
return df
def roc(apply_price, long_period, short_period, u_border, l_border, u_band, l_band):
"""
Parameters
----------
apply_price:价格序列
long_period:长期roc的周期
short_period:短期roc的周期
u_border:长期roc的上界
l_border:长期roc的下界
u_band:均衡线上轨的宽度
l_band:均衡线下轨的宽度
"""
long_roc = talib.ROC(apply_price, timeperiod=long_period)
short_roc = talib.ROC(apply_price, timeperiod=short_period)
_signal_pos = []
u_equilibrium = 100 + u_band
l_equilibrium = 100 - l_band
for index in range(len(apply_price)):
# 卖出点:当长期ROC大于上界且短期ROC在均衡线附近
if long_roc[index] > u_border and (l_equilibrium <= short_roc[index] <= u_equilibrium):
_signal_pos.append((index, 1))
continue
# 买入点:当长期ROC小于下界且短期ROC在均衡线附近
if long_roc[index] < l_border and (l_equilibrium <= short_roc[index] <= u_equilibrium):
_signal_pos.append((index, -1))
continue
return _signal_pos
def cc(candles: np.ndarray,
wma_period: int = 10,
roc_short_period: int = 11,
roc_long_period: int = 14,
source_type: str = "close",
sequential: bool = False) -> Union[float, np.ndarray]:
"""
CC - Coppock Curve
:param candles: np.ndarray
:param wma_period: int - default: 10
:param roc_short_period: int - default: 11
:param roc_long_period: int - default: 14
:param source_type: str - default: "close"
:param sequential: bool - default=False
:return: float | np.ndarray
"""
if not sequential and len(candles) > 240:
candles = candles[-240:]
source = get_candle_source(candles, source_type=source_type)
res = talib.WMA(talib.ROC(source, timeperiod=roc_long_period) +
talib.ROC(source, timeperiod=roc_short_period),
timeperiod=wma_period)
return res if sequential else res[-1]
def _compute_states(self):
start = self.current_idx - self.lookback
end = self.current_idx
df = self._data.iloc[start:end, :]
states = {}
states['roc1'] = ta.ROC(df.close, timeperiod=5).iloc[-1]
states['roc2'] = ta.ROC(df.close, timeperiod=20).iloc[-1]
states['rsi1'] = ta.RSI(df.close, timeperiod=14).iloc[-1]
states['rsi2'] = ta.RSI(df.close, timeperiod=28).iloc[-1]
states['cci1'] = ta.CCI(df.high, df.low, df.close,
timeperiod=14).iloc[-1]
states['cci2'] = ta.CCI(df.high, df.low, df.close,
timeperiod=28).iloc[-1]
states['adx1'] = ta.ADX(df.high, df.low, df.close,
timeperiod=14).iloc[-1]
states['adx2'] = ta.ADX(df.high, df.low, df.close,
timeperiod=28).iloc[-1]
states['atr'] = ta.ATR(df.high, df.low, df.close,
timeperiod=14).iloc[-1]
states['aroon1'] = ta.AROONOSC(df.high, df.low, timeperiod=14).iloc[-1]
states['aroon2'] = ta.AROONOSC(df.high, df.low, timeperiod=28).iloc[-1]
states['bop'] = ta.BOP(df.open, df.high, df.low, df.close).iloc[-1]
states['cmo'] = ta.CMO(df.close, timeperiod=14).iloc[-1]
states['close'] = df.close.iloc[-1]
states['date'] = df.index[-1]
states['position'] = self.position
states['entry_price'] = self.entry_price
states['entry_date'] = self.entry_date
return states
def computeDROC(self):
sroc = talib.ROC(self.prices, timeperiod=self.slowPeriod)
froc = talib.ROC(self.prices, timeperiod=self.fastPeriod)
droc = froc - sroc
droc = sroc
self.removeNullID(droc)
self.rawFeatures['ROC'] = droc
return
def cor(self, sym1, sym2, frequency, period=10):
if not self.kbars_ready(sym1, frequency) or not self.kbars_ready(
sym2, frequency):
return []
close1 = self.close(sym1, frequency)
close2 = self.close(sym2, frequency)
roc1 = ta.ROC(close1, timeperiod=period)
roc2 = ta.ROC(close2, timeperiod=period)
return ta.CORREL(roc1, roc2, timeperiod=period)
def kst(candles: np.ndarray,
sma_period1: int = 10,
sma_period2: int = 10,
sma_period3: int = 10,
sma_period4: int = 15,
roc_period1: int = 10,
roc_period2: int = 15,
roc_period3: int = 20,
roc_period4: int = 30,
signal_period: int = 9,
source_type: str = "close",
sequential: bool = False) -> KST:
"""
Know Sure Thing (KST)
:param candles: np.ndarray
:param sma_period1: int - default=10
:param sma_period2: int - default=10
:param sma_period3: int - default=10
:param sma_period4: int - default=15
:param roc_period1: int - default=10
:param roc_period2: int - default=15
:param roc_period3: int - default=20
:param roc_period4: int - default=30
:param signal_period: int - default=9
:param source_type: str - default: "close"
:param sequential: bool - default=False
:return: KST(line, signal)
"""
warmup_candles_num = get_config('env.data.warmup_candles_num', 240)
if not sequential and len(candles) > warmup_candles_num:
candles = candles[-warmup_candles_num:]
source = get_candle_source(candles, source_type=source_type)
aroc1 = talib.SMA(talib.ROC(source, timeperiod=roc_period1), sma_period1)
aroc2 = talib.SMA(talib.ROC(source, timeperiod=roc_period2), sma_period2)
aroc3 = talib.SMA(talib.ROC(source, timeperiod=roc_period3), sma_period3)
aroc4 = talib.SMA(talib.ROC(source, timeperiod=roc_period4), sma_period4)
line = aroc1[len(aroc1) - len(aroc4):] + 2 * aroc2[len(aroc2) - len(aroc4):] + \
3 * aroc3[len(aroc3) - len(aroc4):] + 4 * aroc4
signal = talib.SMA(line, signal_period)
if sequential:
return KST(line, signal)
else:
return KST(line[-1], signal[-1])
def addTechnicalFeatures(df):
df['Data'] = pd.to_datetime(df['Data'])
df.columns = ['Data', 'Open', 'High', 'Low', 'Close', 'Wol']
#Wskazniki analizy technicznej
k, dfast = talib.STOCH(np.array(df['High']), np.array(
df['Low']), np.array(df['Close'])) # uses high, low, close by default
df['k'] = k
df['dfast'] = dfast
df['dslow'] = talib.SMA(dfast, timeperiod=5)
df['momentum'] = talib.MOM(np.array(df['Close']), timeperiod=4)
df['roc'] = talib.ROC(np.array(df['Close']), timeperiod=5)
df['willR'] = talib.WILLR(np.array(df['High']),
np.array(df['Low']),
np.array(df['Close']),
timeperiod=5)
#ad = talib.ADOSC(np.array(df['High']), np.array(df['Low']),
# np.array(df['Close']), np.array(df['Wol']))
df['disp5'] = df['Close'] / talib.SMA(np.array(df['Close']), 5) * 100
df['disp10'] = df['Close'] / talib.SMA(np.array(df['Close']), 10) * 100
df['oscp'] = ((talib.SMA(np.array(df['Close']), 5) -
talib.SMA(np.array(df['Close']), 10)) /
talib.SMA(np.array(df['Close']), 5))
df['rsi'] = talib.RSI(np.array(df['Close']))
df['CCI'] = talib.CCI(np.array(df['High']), np.array(df['Low']),
np.array(df['Close']))
#Tworzenie zmiennej celu
df['target1'] = df['Close'].shift(-1) - df['Open']
df['target5'] = df['Close'].shift(-5) - df['Open']
df['target3'] = df['Close'].shift(-3) - df['Open']
#zostawienie tylko zmiennej celu i wskaznikow technicznych
df.drop(['Data', 'Open', 'High', 'Low', 'Close', 'Wol'],
axis=1,
inplace=True)
return df
def hsmadata(self, llen, slen): #code 只能是商品
hsmadata = pd.DataFrame()
for code in self.hsmaall.code.unique():
hsma0 = self.hsmaall[self.hsmaall.code == code].copy()
hsma0['ratio'] = 0
hsma0['roc'] = talib.ROC(hsma0.close.values, timeperiod=slen)
hsma0['dayr'] = hsma0.close / hsma0.open - 1
hsma0['dayh'] = hsma0.close / hsma0.high - 1
hsma0['dayl'] = hsma0.close / hsma0.low - 1
hsma0['dayhl'] = hsma0.high / hsma0.low - 1
hsma0['smar'] = hsma0['close'] / talib.SMA(hsma0.close.values,
timeperiod=llen) - 1
hsma0['VAR'] = talib.STDDEV(hsma0.close.values,
timeperiod=llen) / hsma0['close'] - 1
hsma0['timepos'] = 0
for d in hsma0['date'].unique():
hsmad = hsma0[hsma0.date == d]
hsma0.loc[hsma0.date == d, 'ratio'] = (
hsmad.close.iloc[hsmad.shape[0] - 1] / hsmad.close -
1).values
hsma0.loc[hsma0.date == d, 'timepos'] = (
(pd.Series(range(hsmad.shape[0])) + 1) /
hsmad.shape[0]).values
hsma0 = hsma0.dropna()
hsmadata = pd.concat([hsmadata, hsma0], ignore_index=True)
return hsmadata
def kst(prices,
roc1=10,
roc2=15,
roc3=20,
roc4=30,
sma1=10,
sma2=10,
sma3=10,
sma4=15,
smasig=9):
rcma1 = ta.SMA(ta.ROC(prices, timeperiod=roc1), timeperiod=sma1)
rcma2 = ta.SMA(ta.ROC(prices, timeperiod=roc2), timeperiod=sma2)
rcma3 = ta.SMA(ta.ROC(prices, timeperiod=roc3), timeperiod=sma3)
rcma4 = ta.SMA(ta.ROC(prices, timeperiod=roc4), timeperiod=sma4)
kst = 1.0 * rcma1 + 2.0 * rcma2 + 3.0 * rcma3 + 4.0 * rcma4
return ta.SMA(kst, timeperiod=smasig)
def momentum(self): adx = talib.ADX(self.high,self.low,self.close,self.period) adxr = talib.ADXR(self.high,self.low,self.close,self.period) apo = talib.APO(self.high,self.low,self.close,self.period) aroondown, aroonup = talib.AROON(self.high, self.low, period) aroonosc = talib.AROONOSC(self.high,self.low,self.period) bop = talib.BOP(self.opens,self.high,self.low,self.close) cci = talib.CCI(self.high,self.low,self.close,self.period) cmo = talib.CMO(self.close,self.period) dx = talib.DX(self.high,self.low,self.close,self.period) macd, macdsignal, macdhist = talib.MACD(self.close, fastperiod=period, slowperiod=period*5, signalperiod=period*2) macd1, macdsignal1, macdhist1 = talib.MACDEXT(self.close, fastperiod=12, fastmatype=0, slowperiod=26, slowmatype=0, signalperiod=9, signalmatype=0) macd2, macdsignal2, macdhist2 = talib.MACDFIX(self.close, signalperiod=9) mfi = talib.MFI(self.high, self.low, self.close, self.volume, timeperiod=14) minus_di = talib.MINUS_DI(self.high, self.low, self.close, timeperiod=14) minus_dm = talib.MINUS_DM(self.high, self.low, timeperiod=14) mom = talib.MOM(self.close, timeperiod=10) plus_di = talib.PLUS_DI(self.high, self.low, self.close, timeperiod=14) plus_dm = talib.PLUS_DM(self.high, self.low, timeperiod=14) ppo = talib.PPO(self.close, fastperiod=12, slowperiod=26, matype=0) roc = talib.ROC(self.close, timeperiod=10) rocp = talib.ROCP(self.close, timeperiod=10) rocr = talib.ROCR(self.close, timeperiod=10) rocr100 = talib.ROCR100(self.close, timeperiod=10) rsi = talib.RSI(self.close, timeperiod=14) slowk, slowd = talib.STOCH(self.high, self.low, self.close, fastk_period=5, slowk_period=3, slowk_matype=0, slowd_period=3, slowd_matype=0) fastk, fastd = talib.STOCHF(self.high, self.low, self.close, fastk_period=5, fastd_period=3, fastd_matype=0) fastk1, fastd1 = talib.STOCHRSI(self.close, timeperiod=14, fastk_period=5, fastd_period=3, fastd_matype=0) trix = talib.TRIX(self.close, timeperiod=30) ultosc = talib.ULTOSC(self.high, self.low, self.close, timeperiod1=7, timeperiod2=14, timeperiod3=28) willr = talib.WILLR(self.high, self.low, self.close, timeperiod=14)
def get_feature_df(df, company_symbol=None, feature="Trend"):
if (company_symbol != None):
df_comp = df[df["symbol"] == company_symbol]
else:
df_comp = df
df_comp = df_comp.sort_values(by=['time'])
if (feature == "Trend"):
df_comp[feature] = ta.EMA(df_comp.close, timeperiod=90)
elif (feature == "Volatility"):
df_comp[feature] = ta.ATR(df_comp.high,
df_comp.low,
df_comp.close,
timeperiod=90)
elif (feature == "Volume"):
df_comp[feature] = ta.ADOSC(df_comp.high,
df_comp.low,
df_comp.close,
df_comp.volume,
fastperiod=30,
slowperiod=90)
elif (feature == "Momemtum"):
df_comp[feature + "_1"] = ta.RSI(df_comp.close, timeperiod=90)
df_comp[feature + "_2"] = ta.ROC(df_comp.close, timeperiod=90)
elif (feature == "Open_Close"):
df_comp[feature + "_1"] = df_comp["open"]
df_comp[feature + "_2"] = df_comp["close"]
return df_comp
def compROC(self):
roc = talib.ROC(self.close,timeperiod=self.lookback)
self.removeNullID(roc)
self.rawFeatures['ROC'] = roc
FEATURE_SIZE_DICT['ROC'] = 1
return
def get_tis(df, dropnan=True, drop_vol=True):
""" calculate technical indicators with ta-lib
requires a df with a column named "Close" """
# to delete rows which have 0 volume / delete weekends
print("dropping rows with zero volume from dataFrame {}".format(df.shape))
if drop_vol:
df = df[(df[['Volume']] != 0).all(axis=1)]
print("creating technical indicators")
df['RSI'] = ta.RSI(df.Close.values, timeperiod=14)
df['ROC'] = ta.ROC(df.Close.values, timeperiod=10)
df['SMA'] = ta.SMA(df.Close.values, timeperiod=30)
df['EMA'] = ta.EMA(df.Close.values, timeperiod=30)
df['WMA'] = ta.WMA(df.Close.values, timeperiod=30)
df['MACD'], df['macdSignal'], df['macdHist'] = ta.MACD(df.Close.values,
fastperiod=12,
slowperiod=26,
signalperiod=9)
print("done {}".format(df.shape))
print("dropping NaN values")
if dropnan:
df.dropna(inplace=True)
print("returning dataFrame: {} from get_tis func".format(df.shape))
return df
def roc(candles: np.ndarray,
period: int = 10,
source_type: str = "close",
sequential: bool = False) -> Union[float, np.ndarray]:
"""
ROC - Rate of change : ((price/prevPrice)-1)*100
:param candles: np.ndarray
:param period: int - default=10
:param source_type: str - default: "close"
:param sequential: bool - default=False
:return: float | np.ndarray
"""
warmup_candles_num = get_config('env.data.warmup_candles_num', 240)
if not sequential and len(candles) > warmup_candles_num:
candles = candles[-warmup_candles_num:]
source = get_candle_source(candles, source_type=source_type)
res = talib.ROC(source, timeperiod=period)
if sequential:
return res
else:
return None if np.isnan(res[-1]) else res[-1]
def test_indicator_ROC(self):
n = 3
price = 'Close'
result = ROC(df, n)
isinstance(result, pd.DataFrame)
expected = talib.ROC(df[price].values, timeperiod=n)
np.testing.assert_almost_equal(result.values, expected)
def calc_rps_returns_roc_func(data, *args, **kwargs):
"""
计算收益率 实现算法2，经测试和算法1无速度差别，结果一致
w:周5;月20;半年：120; 一年250
"""
w = 63
indices = None
# 针对多标的，拆分 indices 数据再自动合并
code = data.index.get_level_values(level=1)[0]
if (len(kwargs.keys()) > 0):
w = kwargs['w'] if ('w' in kwargs.keys()) else w
indices = kwargs['indices'].loc[(slice(None), code), :] if (
'indices' in kwargs.keys()) else None
if (len(args) > 0):
w = args[0] if (len(args) > 0) else w
indices = args[1].loc[(slice(None),
code), :] if (len(args) >= 2) else None
rps_reutrns = talib.ROC(data.close, w) / 100
if (indices is None):
return pd.DataFrame(rps_reutrns,
columns=[FLD.RPS_RETURNS],
index=data.index).fillna(0)
else:
return pd.concat([
indices,
pd.Series(rps_reutrns, name=FLD.RPS_RETURNS,
index=data.index).fillna(0)
],
axis=1)
def try_bottom_strategy(df):
df['upperband'], df['middleband'], df['lowerband'] = talib.BBANDS(
df['close'], timeperiod=30, nbdevup=3, nbdevdn=3, matype=0)
df['std_var'] = (df['upperband'] - df['middleband']) / 3
df['close_off'] = (df['close'] - df['middleband']) / df['std_var']
df['rsi_14'] = talib.RSI(df['close'], timeperiod=14)
df['vol_diff'] = talib.ROC(df['volume'], timeperiod=1)
df['true_range'] = (df['high'] - df['low']) / df['low'] * 100
df['true_bar'] = (df['close'] - df['open']) / df['open'] * 100
df = df.round(2)
df = df.dropna(how='any')
df = df.reset_index(drop=True)
score_ls = []
for i in range(len(df)):
score = 0
if df['close_off'][i] < -1.6:
score = score + 1
if df['rsi_14'][i] < 50:
score = score + 1
if df['vol_diff'][i] > 80:
score = score + 1
if df['true_range'][i] < 5:
score = score + 1
if abs(df['true_bar'][i]) < 3:
score = score + 1
score_ls.append(score)
df['score'] = score_ls
return df
def get_features(data):
tech_data = pd.DataFrame(index=data.index);
for t in periods_list:
tech_data[f'SMA_{t}'] = talib.SMA(data.close,timeperiod=t)
tech_data[f'MOM_{t}'] = talib.MOM(data.close, timeperiod=t)
tech_data[f'RSI_{t}'] = talib.RSI(data.close, timeperiod=t)
tech_data[f'MA_{t}'] = talib.MA(data.close, timeperiod=t)
tech_data[f'DX_{t}'] = talib.DX(data.high, data.low, data.close, timeperiod=t)
tech_data[f'volume_change_{t}'] = data.volume.pct_change(periods=t)
tech_data[f'volatility_{t}'] = data.close.pct_change(periods=t).std()
tech_data[f'ADX_{t}'] = talib.ADX(data.high, data.low, data.close, timeperiod=t)
tech_data[f'ADXR_{t}'] = talib.ADXR(data.high, data.low, data.close, timeperiod=t)
tech_data[f'AROONOSC_{t}'] = talib.AROONOSC(data.high, data.low, timeperiod=t)
tech_data[f'ROC_{t}'] = talib.ROC(data.close, timeperiod=t)
tech_data[f'BIAS_{t}'] = (data['close'] - data['close'].rolling(t, min_periods=1).mean())/ data['close'].rolling(t, min_periods=1).mean()*100
tech_data[f'BOLL_upper_{t}'], tech_data[f'BOLL_middle_{t}'], tech_data[f'BOLL_lower_{t}'] = talib.BBANDS(
data.close,
timeperiod=t,
nbdevup=2,
nbdevdn=2,
matype=0)
tech_data['SAR'] = talib.SAR(data.high, data.low)
tech_data['AD'] = talib.AD(data.high, data.low, data.close, data.volume)
tech_data['OBV'] = talib.OBV(data.close, data.volume)
tech_data['target'] = data.close.pct_change().shift(-1).apply(lambda x: 1 if x > 0 else -1).fillna(0)
tech_data['time'] = data.time
tech_data = tech_data.set_index('time')
reduce(lambda x, y: cross_over(x, y, tech_data), periods_list)
features = list(set(tech_data.columns) - set(data.columns) - set(['target']))
return tech_data.dropna(), features
def rebalance(context, data):
history = data.history(assets=context.asset,
fields=['close'],
bar_count=context.roc_window + 1,
frequency='1d')
date = history.index.values[-1]
close = history['close'].values
roc = ta.ROC(close, timeperiod=context.roc_window)
current_price = data[context.asset].price
record(price=current_price)
buy_signal_triggered = False
sell_signal_triggered = False
if roc[-1] > 0:
buy_signal_triggered = True
elif roc[-1] < 0:
sell_signal_triggered = True
current_position = context.portfolio.positions[context.asset].amount
if buy_signal_triggered and current_position == 0:
print(str(date) + '==>Buy')
order_target_percent(context.asset, 1.0)
elif sell_signal_triggered and current_position > 0:
print(str(date) + '==>Sell')
order_target_percent(context.asset, 0.0)
else:
print("No trading")
def test_roc():
"""test TA.ROC"""
roc = TA.ROC(ohlc, 10)
talib_roc = talib.ROC(ohlc["close"], 10)
assert round(talib_roc[-1], 5) == round(roc.values[-1], 5)
def genTA(data, y, t): #t is timeperiod
indicators = {}
y_ind = copy.deepcopy(y)
for ticker in data:
## Overlap
indicators[ticker] = ta.SMA(data[ticker].iloc[:,3], timeperiod=t).to_frame()
indicators[ticker]['EMA'] = ta.EMA(data[ticker].iloc[:,3], timeperiod=t)
indicators[ticker]['BBAND_Upper'], indicators[ticker]['BBAND_Middle' ], indicators[ticker]['BBAND_Lower' ] = ta.BBANDS(data[ticker].iloc[:,3], timeperiod=t, nbdevup=2, nbdevdn=2, matype=0)
indicators[ticker]['HT_TRENDLINE'] = ta.HT_TRENDLINE(data[ticker].iloc[:,3])
indicators[ticker]['SAR'] = ta.SAR(data[ticker].iloc[:,1], data[ticker].iloc[:,2], acceleration=0, maximum=0)
#rename SMA column
indicators[ticker].rename(columns={indicators[ticker].columns[0]: "SMA"}, inplace=True)
## Momentum
indicators[ticker]['RSI'] = ta.RSI(data[ticker].iloc[:,3], timeperiod=(t-1))
indicators[ticker]['MOM'] = ta.MOM(data[ticker].iloc[:,3], timeperiod=(t-1))
indicators[ticker]['ROC'] = ta.ROC(data[ticker].iloc[:,3], timeperiod=(t-1))
indicators[ticker]['ROCP']= ta.ROCP(data[ticker].iloc[:,3],timeperiod=(t-1))
indicators[ticker]['STOCH_SLOWK'], indicators[ticker]['STOCH_SLOWD'] = ta.STOCH(data[ticker].iloc[:,1], data[ticker].iloc[:,2], data[ticker].iloc[:,3], fastk_period=t, slowk_period=int(.6*t), slowk_matype=0, slowd_period=int(.6*t), slowd_matype=0)
indicators[ticker]['MACD'], indicators[ticker]['MACDSIGNAL'], indicators[ticker]['MACDHIST'] = ta.MACD(data[ticker].iloc[:,3], fastperiod=t,slowperiod=2*t,signalperiod=int(.7*t))
## Volume
indicators[ticker]['OBV'] = ta.OBV(data[ticker].iloc[:,3], data[ticker].iloc[:,4])
indicators[ticker]['AD'] = ta.AD(data[ticker].iloc[:,1], data[ticker].iloc[:,2], data[ticker].iloc[:,3], data[ticker].iloc[:,4])
indicators[ticker]['ADOSC'] = ta.ADOSC(data[ticker].iloc[:,1], data[ticker].iloc[:,2], data[ticker].iloc[:,3], data[ticker].iloc[:,4], fastperiod=int(.3*t), slowperiod=t)
## Cycle
indicators[ticker]['HT_DCPERIOD'] = ta.HT_DCPERIOD(data[ticker].iloc[:,3])
indicators[ticker]['HT_TRENDMODE']= ta.HT_TRENDMODE(data[ticker].iloc[:,3])
## Price
indicators[ticker]['AVGPRICE'] = ta.AVGPRICE(data[ticker].iloc[:,0], data[ticker].iloc[:,1], data[ticker].iloc[:,2], data[ticker].iloc[:,3])
indicators[ticker]['TYPPRICE'] = ta.TYPPRICE(data[ticker].iloc[:,1], data[ticker].iloc[:,2], data[ticker].iloc[:,3])
## Volatility
indicators[ticker]['ATR'] = ta.ATR(data[ticker].iloc[:,1], data[ticker].iloc[:,2], data[ticker].iloc[:,3], timeperiod=(t-1))
## Statistics
indicators[ticker]['BETA'] = ta.BETA(data[ticker].iloc[:,1], data[ticker].iloc[:,2], timeperiod=(t-1))
indicators[ticker]['LINEARREG'] = ta.LINEARREG(data[ticker].iloc[:,3], timeperiod=t)
indicators[ticker]['VAR'] = ta.VAR(data[ticker].iloc[:,3], timeperiod=t, nbdev=1)
## Math Transform
indicators[ticker]['EXP'] = ta.EXP(data[ticker].iloc[:,3])
indicators[ticker]['LN'] = ta.LN(data[ticker].iloc[:,3])
## Patterns (returns integers - but norming might not really do anything but wondering if they should be normed)
indicators[ticker]['CDLENGULFING'] = ta.CDLENGULFING(data[ticker].iloc[:,0], data[ticker].iloc[:,1], data[ticker].iloc[:,2], data[ticker].iloc[:,3])
indicators[ticker]['CDLDOJI'] = ta.CDLDOJI(data[ticker].iloc[:,0], data[ticker].iloc[:,1], data[ticker].iloc[:,2], data[ticker].iloc[:,3])
indicators[ticker]['CDLHAMMER'] = ta.CDLHAMMER(data[ticker].iloc[:,0], data[ticker].iloc[:,1], data[ticker].iloc[:,2], data[ticker].iloc[:,3])
indicators[ticker]['CDLHANGINGMAN']= ta.CDLHANGINGMAN(data[ticker].iloc[:,0], data[ticker].iloc[:,1], data[ticker].iloc[:,2], data[ticker].iloc[:,3])
#drop 'nan' values
indicators[ticker].drop(indicators[ticker].index[np.arange(0,63)], inplace=True)
y_ind[ticker].drop(y_ind[ticker].index[np.arange(0,63)], inplace=True)
#Normalize Features
indicators_norm = normData(indicators)
return indicators_norm, indicators, y_ind
def calculate_roc(self, period=1):
"""
Rate of change : ((price/prevPrice)-1)*100
:return:
"""
real = talib.ROC(self.close, timeperiod=period)
return real
def technical_indicators(ohlcv_path="data/OHLC_NDX.csv"):
"""
Calculate Technical Indicators from the ti_dict dictionary and export to .csv
:param ohlcv_path: OHLCV dataset path
"""
ohlcv = pd.read_csv(ohlcv_path)
ti_dict = {
'SMA5': talib.SMA(ohlcv.Close, 5),
'SMA10': talib.SMA(ohlcv.Close, 10),
'SMA50': talib.SMA(ohlcv.Close, 50),
'EMA20': talib.EMA(ohlcv.Close, 20),
'stoch5': talib.STOCH(ohlcv.High, ohlcv.Low, ohlcv.Close, 5, 3, 0, 3, 0)[0],
'ADOSC': talib.ADOSC(ohlcv.High, ohlcv.Low, ohlcv.Close, ohlcv.Volume, fastperiod=3, slowperiod=10),
'MACDhist': talib.MACD(ohlcv.Close, fastperiod=12, slowperiod=26, signalperiod=9)[2],
'WILLR': talib.WILLR(ohlcv.High, ohlcv.Low, ohlcv.Close, timeperiod=14),
'RSI': talib.RSI(ohlcv.Close, timeperiod=14),
'MOM': talib.MOM(ohlcv.Close, timeperiod=10),
'ROC': talib.ROC(ohlcv.Close, timeperiod=10),
'OBV': talib.OBV(ohlcv.Close, ohlcv.Volume),
'CCI': talib.CCI(ohlcv.High, ohlcv.Low, ohlcv.Close, timeperiod=14)
}
# Create a dataframe from TI dictionary
df_ti = pd.DataFrame(ti_dict, index=ohlcv.index)
del ti_dict
del ohlcv
# Save Technical Indicators dataframe
if not os.path.isdir('data_preprocessed/'):
os.mkdir('data_preprocessed/')
df_ti.to_csv('data_preprocessed/technical_indicators.csv', index=False)
del df_ti
return 1
def generate_feature(data):
high = data.High.values
low = data.Low.values
close = data.Close.values
feature_df = pd.DataFrame(index=data.index)
feature_df["ADX"] = ADX = talib.ADX(high, low, close, timeperiod=14)
feature_df["ADXR"] = ADXR = talib.ADXR(high, low, close, timeperiod=14)
feature_df["APO"] = APO = talib.APO(close, fastperiod=12, slowperiod=26, matype=0)
feature_df["AROONOSC"] = AROONOSC = talib.AROONOSC(high, low, timeperiod=14)
feature_df["CCI"] = CCI = talib.CCI(high, low, close, timeperiod=14)
feature_df["CMO"] = CMO = talib.CMO(close, timeperiod=14)
feature_df["DX"] = DX = talib.DX(high, low, close, timeperiod=14)
feature_df["MINUS_DI"] = MINUS_DI = talib.MINUS_DI(high, low, close, timeperiod=14)
feature_df["MINUS_DM"] = MINUS_DM = talib.MINUS_DM(high, low, timeperiod=14)
feature_df["MOM"] = MOM = talib.MOM(close, timeperiod=10)
feature_df["PLUS_DI"] = PLUS_DI = talib.PLUS_DI(high, low, close, timeperiod=14)
feature_df["PLUS_DM"] = PLUS_DM = talib.PLUS_DM(high, low, timeperiod=14)
feature_df["PPO"] = PPO = talib.PPO(close, fastperiod=12, slowperiod=26, matype=0)
feature_df["ROC"] = ROC = talib.ROC(close, timeperiod=10)
feature_df["ROCP"] = ROCP = talib.ROCP(close, timeperiod=10)
feature_df["ROCR100"] = ROCR100 = talib.ROCR100(close, timeperiod=10)
feature_df["RSI"] = RSI = talib.RSI(close, timeperiod=14)
feature_df["ULTOSC"] = ULTOSC = talib.ULTOSC(high, low, close, timeperiod1=7, timeperiod2=14, timeperiod3=28)
feature_df["WILLR"] = WILLR = talib.WILLR(high, low, close, timeperiod=14)
feature_df = feature_df.fillna(0.0)
matrix = np.stack((
ADX, ADXR, APO, AROONOSC, CCI, CMO, DX, MINUS_DI, ROCR100, ROC,
MINUS_DM, MOM, PLUS_DI, PLUS_DM, PPO, ROCP, WILLR, ULTOSC, RSI))
matrix = np.nan_to_num(matrix)
matrix = matrix.transpose()
return feature_df, matrix
def getData(code,start,end):
data = pandasData.DataReader(code,'yahoo',start,end)
# print len(data)
f_ema = ta.EMA(data['Close'].values, timeperiod=30).tolist()
# print f_ema
f_ma = ta.MA(data['Close'].values, timeperiod=30, matype=0).tolist()
f_wma = ta.WMA(data['Close'].values, timeperiod=30).tolist()
f_momentum = ta.MOM(data['Close'].values, timeperiod=10).tolist()
f_roc = ta.ROC(data['Close'].values, timeperiod=10).tolist()
# f_cycle = ta.HT_DCPERIOD(data['Close'].values).tolist()
f_price = ta.WCLPRICE(data['High'].values, data['Low'].values, data['Close'].values).tolist()
f_natr = ta.NATR(data['High'].values, data['Low'].values, data['Close'].values, timeperiod=14).tolist()
f_stddev = ta.STDDEV(data['Close'].values, timeperiod=5, nbdev=1).tolist()
X = pd.DataFrame(
pd.np.array([f_ema, f_ma, f_wma, f_momentum, f_roc, f_price, f_natr, f_stddev]).T[32:]
,columns=['f_ema','f_ma','f_wma','f_momentum','f_roc','f_price','f_natr','f_stddev'])
# print X['f_ema'].size
# print X
data = data['Close'].tolist()
finaldata = [[] for i in range(2)]
for i in range(0, len(data) - 1):
temp = (data[i + 1] - data[i]) / data[i]
finaldata[0].append(temp)
if (temp > 0):
finaldata[1].append(1)
else:
finaldata[1].append(0)
# print len(data)
data = data[31:len(data) - 1]
# print data
Y = pd.DataFrame(pd.np.array(finaldata).T, columns=['change', 'label'])
X = X.join(Y)
return X
def getMomentumIndicators(df):
high = df['High']
low = df['Low']
close = df['Close']
open = df['Open']
volume = df['Volume']
df['ADX'] = ta.ADX(high, low, close, timeperiod=14)
df['SMA'] = ta.ADXR(high, low, close, timeperiod=14)
df['APO'] = ta.APO(close, fastperiod=12, slowperiod=26, matype=0)
df['AROONDOWN'], df['AROOONUP'] = ta.AROON(high, low, timeperiod=14)
df['AROONOSC'] = ta.AROONOSC(high, low, timeperiod=14)
df['BOP'] = ta.BOP(open, high, low, close)
df['CCI'] = ta.CCI(high, low, close, timeperiod=14)
df['CMO'] = ta.CMO(close, timeperiod=14)
df['DX'] = ta.DX(high, low, close, timeperiod=14)
df['MACD'], df['MACDSIGNAL'], df['MACDHIST'] = ta.MACD(close, fastperiod=12, slowperiod=26, signalperiod=9)
df['MFI'] = ta.MFI(high, low, close, volume, timeperiod=14)
df['MINUS_DI'] = ta.MINUS_DI(high, low, close, timeperiod=14)
df['MINUS_DM']= ta.MINUS_DM(high, low, timeperiod=14)
df['MOM'] = ta.MOM(close, timeperiod=10)
df['PLUS_DM'] =ta.PLUS_DM(high, low, timeperiod=14)
df['PPO'] = ta.PPO(close, fastperiod=12, slowperiod=26, matype=0)
df['ROC'] = ta.ROC(close, timeperiod=10)
df['ROCP'] = ta.ROCP(close, timeperiod=10)
df['ROCR'] = ta.ROCR(close, timeperiod=10)
df['ROCR100'] = ta.ROCR100(close, timeperiod=10)
df['RSI'] = ta.RSI(close, timeperiod=14)
df['SLOWK'], df['SLOWD'] = ta.STOCH(high, low, close, fastk_period=5, slowk_period=3, slowk_matype=0, slowd_period=3, slowd_matype=0)
df['FASTK'], df['FASTD'] = ta.STOCHF(high, low, close, fastk_period=5, fastd_period=3, fastd_matype=0)
df['FASTK2'], df['FASTD2'] = ta.STOCHRSI(close, timeperiod=14, fastk_period=5, fastd_period=3, fastd_matype=0)
df['TRIX'] = ta.TRIX(close, timeperiod=30)
df['ULTOSC'] = ta.ULTOSC(high, low, close, timeperiod1=7, timeperiod2=14, timeperiod3=28)
df['WILLR'] = ta.WILLR(high, low, close, timeperiod=14)
def rocStretagy(companyname, begin):
end = datetime.date.today()
#begin=end-pd.DateOffset(365*backtestyear)
#timestamp format and get apple stock.
st = dt.strptime(begin, '%Y-%m-%d')
#st=begin.strftime('%Y-%m-%d')
ed = end.strftime('%Y-%m-%d')
symbol = companyname + ".NS"
datanifty = nift50Indices.objects.filter(Ticker=symbol,
Date__gte=st).values_list(
'Close', 'Open', 'High',
'Low')
#data = pdr.get_data_yahoo(company,st,ed)
dataframe = pd.DataFrame(list(datanifty),
columns=['Close', 'Open', 'High', 'Low'])
print(dataframe)
data = dataframe.astype(float)
#data = pdr.get_data_yahoo('RELIANCE.NS',st,ed)
#Rate of change ROC
data['ROC'] = talib.ROC(np.asarray(data['Close']), 20)
graph = data.plot(y=['ROC'], title=companyname + ' ROC ')
graph = plt.gcf()
fig_html = mpld3.fig_to_html(graph)
return fig_html
def backtest_market(data, buy_barrier, short_barrier):
COL_TIME = 0
COL_CLOSE = 4
COL_VOLUME = 5
closes = [x[COL_CLOSE] for x in data]
dates = [x[COL_TIME] for x in data]
volumes = [x[COL_VOLUME] for x in data]
candle_data = {'close': closes, 'volume': volumes}
bars = pd.DataFrame(candle_data, index=dates, columns=['close', 'volume'])
#print (bars.describe())
bars['Pct Change'] = bars['close'].astype('float').pct_change()
bars['RSI'] = talib.RSI(bars['close'])
bars['volumeROC'] = talib.ROC(bars['volume'])
symbol = market
rfs = RSIStrategy('close', bars, buy_barrier, short_barrier)
signals = rfs.generate_signals()
#print ("signal summary \n",signals.describe())
backtest = MarketOnOpenPortfolio(symbol, bars, signals)
pf = backtest.backtest_portfolio()
lastindex = pf['Strategy'][-1]
total_return = lastindex / 100 - 1
print("total return ", total_return)
plot_portfolio(pf)
