def ultosc(self,
time_period1: int = 7,
time_period2: int = 14,
time_period3: int = 28,
array: bool = False) -> Union[float, np.ndarray]:
"""
Ultimate Oscillator.
"""
result = talib.ULTOSC(self.high, self.low, self.close, time_period1,
time_period2, time_period3)
if array:
return result
return result[-1]
def add_tech_ind(self, price_map):
# Visit for technical indicator documentation: http://mrjbq7.github.io/ta-lib/funcs.html
new_price_map = {}
for symbol in price_map:
symbol_data = price_map[symbol]
symbol_data['osc'] = ta.ULTOSC(symbol_data['high'],
symbol_data['low'],
symbol_data['open'], 7, 14, 21)
new_price_map[symbol] = symbol_data
return new_price_map
def test_uo(self):
result = pandas_ta.uo(self.high, self.low, self.close)
self.assertIsInstance(result, Series)
self.assertEqual(result.name, 'UO_7_14_28')
try:
expected = tal.ULTOSC(self.high, self.low, self.close)
pdt.assert_series_equal(result, expected, check_names=False)
except AssertionError as ae:
try:
corr = pandas_ta.utils.df_error_analysis(result, expected, col=CORRELATION)
self.assertGreater(corr, CORRELATION_THRESHOLD)
except Exception as ex:
error_analysis(result, CORRELATION, ex)
def populateindicators(dataframe):
bardata = dataframe.assign(mfi=ta.MFI(dataframe['high'],
dataframe['low'],
dataframe['close'],
np.asarray(dataframe['volume'],
dtype='float'),
timeperiod=14),
ultosc=ta.ULTOSC(dataframe['high'],
dataframe['low'],
dataframe['close']),
minusdi=tab.MINUS_DI(dataframe, timeperiod=14),
plusdi=tab.PLUS_DI(dataframe, timeperiod=14),
rsi=tab.RSI(dataframe,
timeperiod=14,
price='close'),
adx=tab.ADX(dataframe))
return bardata
def ultosc(
client,
symbol,
range="6m",
highcol="high",
lowcol="low",
closecol="close",
period1=7,
period2=14,
period3=28,
):
"""This will return a dataframe of
Ultimate Oscillator
for the given symbol across the given range
Args:
client (pyEX.Client): Client
symbol (string): Ticker
range (string): range to use, for pyEX.chart
highcol (string): column to use to calculate
lowcol (string): column to use to calculate
closecol (string): column to use to calculate
period1 (int): period to calculate across
period2 (int): period to calculate across
period3 (int): period to calculate across
Returns:
DataFrame: result
"""
df = client.chartDF(symbol, range)
x = t.ULTOSC(
df[highcol].values.astype(float),
df[lowcol].values.astype(float),
df[closecol].values.astype(float),
timeperiod1=period1,
timeperiod2=period2,
timeperiod3=period3,
)
return pd.DataFrame(
{
highcol: df[highcol].values,
lowcol: df[lowcol].values,
closecol: df[closecol].values,
"ultosc": x,
}
)
def generate_tech_data_default(stock,
open_name,
close_name,
high_name,
low_name,
volume_name='vol'):
open_price = stock[open_name].values
close_price = stock[close_name].values
low_price = stock[low_name].values
high_price = stock[high_name].values
volume = stock[volume_name].values
data = stock.copy()
data['MOM'] = talib.MOM(close_price)
data['HT_DCPERIOD'] = talib.HT_DCPERIOD(close_price)
data['HT_DCPHASE'] = talib.HT_DCPHASE(close_price)
data['sine'], data['leadsine'] = talib.HT_SINE(close_price)
data['inphase'], data['quadrature'] = talib.HT_PHASOR(close_price)
data['ADXR'] = talib.ADXR(high_price, low_price, close_price)
data['APO'] = talib.APO(close_price)
data['AROON_UP'], _ = talib.AROON(high_price, low_price)
data['CCI'] = talib.CCI(high_price, low_price, close_price)
data['PLUS_DI'] = talib.PLUS_DI(high_price, low_price, close_price)
data['PPO'] = talib.PPO(close_price)
data['macd'], data['macd_sig'], data['macd_hist'] = talib.MACD(close_price)
data['CMO'] = talib.CMO(close_price)
data['ROCP'] = talib.ROCP(close_price)
data['fastk'], data['fastd'] = talib.STOCHF(high_price, low_price,
close_price)
data['TRIX'] = talib.TRIX(close_price)
data['ULTOSC'] = talib.ULTOSC(high_price, low_price, close_price)
data['WILLR'] = talib.WILLR(high_price, low_price, close_price)
data['NATR'] = talib.NATR(high_price, low_price, close_price)
data['MFI'] = talib.MFI(high_price, low_price, close_price, volume)
data['RSI'] = talib.RSI(close_price)
data['AD'] = talib.AD(high_price, low_price, close_price, volume)
data['OBV'] = talib.OBV(close_price, volume)
data['EMA'] = talib.EMA(close_price)
data['SAREXT'] = talib.SAREXT(high_price, low_price)
data['TEMA'] = talib.EMA(close_price)
#data = data.drop([open_name, close_name, high_name, low_name, volume_name, 'amount', 'count'], axis=1)
data = drop_columns(data, [
open_name, close_name, high_name, low_name, volume_name, 'amount',
'count'
])
data = data.dropna().astype(np.float32)
return data
def getCustomData(df):
open = df['Open']
high = df['High']
low = df['Low']
close = df['Close']
volume = df['Volume']
df['TRENDMODE'] = ta.HT_TRENDMODE(close)
df['RSI'] = ta.RSI(close,timeperiod=14)
df['RSI2'] = ta.RSI(close,timeperiod=7)
df['RSI3'] = ta.RSI(close,timeperiod=28)
df['ADX1'] = ta.ADX(high,low,close,timeperiod=7)
df['ADX2'] = ta.ADX(high,low,close,timeperiod=28)
df['ADX'] = ta.ADX(high,low,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['ULTOSC'] = ta.ULTOSC(high, low, close, timeperiod1=7, timeperiod2=14, timeperiod3=28)
df['WILLR'] = ta.WILLR(high, low, close, timeperiod=14)
def calc_indicators(df):
po = df.open.values
ph = df.high.values
pl = df.low.values
pc = df.close.values
po = np.array(po, dtype=float)
ph = np.array(ph, dtype=float)
pl = np.array(pl, dtype=float)
pc = np.array(pc, dtype=float)
ma = []
ema = []
for i in [5, 10, 20, 60, 120]:
ma.append(talib.MA(pc, i)[-1])
ema.append(talib.EMA(pc, i)[-1])
macd = talib.MACD(pc)[2][-1]
rsi = talib.RSI(pc, 14)[-1]
stoch = talib.STOCH(ph, pl, pc, 9, 6)[0][-1]
adx = talib.ADX(ph, pl, pc, 14)[-1]
cci = talib.CCI(ph, pl, pc, 14)[-1]
# willr = talib.WILLR(ph, pl, pc, 14)[-1]
stochrsi = talib.STOCHRSI(pc, 14)[0][-1]
uo = talib.ULTOSC(ph, pl, pc)[-1]
roc = pc[-1] / pc[-2] - 1
sar = talib.SAR(ph, pl)[-1]
plusDI = talib.PLUS_DI(ph, pl, pc, 14)[-1]
minusDI = talib.MINUS_DI(ph, pl, pc, 14)[-1]
adx *= (lambda x: 1 if x > 0 else -1 if x < 0 else 0)(plusDI - minusDI)
# print(ma, ema, macd, rsi, stoch, adx, cci, stochrsi, uo, roc, sar)
indicators = {'ma': ma,
'ema': ema,
'macd': macd,
'rsi': rsi,
'stoch': stoch,
'adx': adx,
'cci': cci,
'stochrsi': stochrsi,
'uo': uo,
'roc': roc,
'sar': sar}
return indicators
def updateIndicator(self, data, candles_visible):
# ULTOSC(high, low, close, timeperiod1=7, timeperiod2=14, timeperiod3=28)
ult_osc = talib.ULTOSC(data[2],
data[3],
data[4],
timeperiod1=7,
timeperiod2=14,
timeperiod3=28)
try:
firstNotNan = np.where(np.isnan(ult_osc))[0][-1] + 1
except:
firstNotNan = 0
ult_osc[:firstNotNan] = ult_osc[firstNotNan]
ult_osc = ult_osc[-candles_visible:]
self.line.clear()
for i in range(candles_visible):
self.line.append(i + 0.5, ult_osc[i])
# detach and remove old axes
for ax in self.line.attachedAxes():
self.line.detachAxis(ax)
for ax in self.axes():
self.removeAxis(ax)
# set x axes
ax = QtChart.QValueAxis()
ax.setRange(0, candles_visible)
ax.hide()
# set y axes
ay = QtChart.QValueAxis()
ay.setRange(0, 100)
ay.setGridLinePen(QtGui.QPen(QtGui.QColor(80, 80, 80), 0.5))
ay.setLinePen(QtGui.QPen(QtGui.QColor(0, 0, 0), 0.5))
ay.applyNiceNumbers()
ay.setTickCount(3)
# add and attach new axes
self.addAxis(ax, QtCore.Qt.AlignBottom)
self.addAxis(ay, QtCore.Qt.AlignRight)
self.line.attachAxis(ax)
self.line.attachAxis(ay)
def generate_tech_data(stock, open_name, close_name, high_name, low_name):
open_price = stock[open_name].values
close_price = stock[close_name].values
low_price = stock[low_name].values
high_price = stock[high_name].values
data = pd.DataFrame(stock)
data['MOM'] = talib.MOM(close_price)
data['SMA'] = talib.SMA(close_price)
data['HT_DCPERIOD'] = talib.HT_DCPERIOD(close_price)
data['HT_DCPHASE'] = talib.HT_DCPHASE(close_price)
data['sine'], data['leadsine'] = talib.HT_SINE(close_price)
data['inphase'], data['quadrature'] = talib.HT_PHASOR(close_price)
data['HT_TRENDMODE'] = talib.HT_TRENDMODE(close_price)
data['SAREXT'] = talib.SAREXT(high_price, low_price)
data['ADX'] = talib.ADX(high_price, low_price, close_price)
data['ADXR'] = talib.ADX(high_price, low_price, close_price)
data['APO'] = talib.APO(close_price)
data['AROON_UP'], data['AROON_DOWN'] = talib.AROON(high_price, low_price)
data['AROONOSC'] = talib.AROONOSC(high_price, low_price)
data['BOP'] = talib.BOP(open_price, high_price, low_price, close_price)
data['CCI'] = talib.CCI(high_price, low_price, close_price)
data['PLUS_DI'] = talib.PLUS_DI(high_price, low_price, close_price)
data['PLUS_DM'] = talib.PLUS_DM(high_price, low_price)
data['PPO'] = talib.PPO(close_price)
data['macd'], data['macd_sig'], data['macd_hist'] = talib.MACD(close_price)
data['RSI'] = talib.RSI(close_price)
data['CMO'] = talib.CMO(close_price)
data['ROC'] = talib.ROC(close_price)
data['ROCP'] = talib.ROCP(close_price)
data['ROCR'] = talib.ROCR(close_price)
data['slowk'], data['slowd'] = talib.STOCH(high_price, low_price,
close_price)
data['fastk'], data['fastd'] = talib.STOCHF(high_price, low_price,
close_price)
data['TRIX'] = talib.TRIX(close_price)
data['ULTOSC'] = talib.ULTOSC(high_price, low_price, close_price)
data['WILLR'] = talib.WILLR(high_price, low_price, close_price)
data['NATR'] = talib.NATR(high_price, low_price, close_price)
data['TRANGE'] = talib.TRANGE(high_price, low_price, close_price)
data = data.drop([open_name, close_name, high_name, low_name], axis=1)
data = data.dropna()
return data
def _get_indicators(security, open_name, close_name, high_name, low_name,
volume_name):
open_price = security[open_name].values
close_price = security[close_name].values
low_price = security[low_name].values
high_price = security[high_name].values
volume = security[volume_name].values if volume_name else None
security['MOM'] = talib.MOM(close_price)
security['HT_DCPERIOD'] = talib.HT_DCPERIOD(close_price)
security['HT_DCPHASE'] = talib.HT_DCPHASE(close_price)
security['SINE'], security['LEADSINE'] = talib.HT_SINE(close_price)
security['INPHASE'], security['QUADRATURE'] = talib.HT_PHASOR(
close_price)
security['ADXR'] = talib.ADXR(high_price, low_price, close_price)
security['APO'] = talib.APO(close_price)
security['AROON_UP'], _ = talib.AROON(high_price, low_price)
security['CCI'] = talib.CCI(high_price, low_price, close_price)
security['PLUS_DI'] = talib.PLUS_DI(high_price, low_price, close_price)
security['PPO'] = talib.PPO(close_price)
security['MACD'], security['MACD_SIG'], security[
'MACD_HIST'] = talib.MACD(close_price)
security['CMO'] = talib.CMO(close_price)
security['ROCP'] = talib.ROCP(close_price)
security['FASTK'], security['FASTD'] = talib.STOCHF(
high_price, low_price, close_price)
security['TRIX'] = talib.TRIX(close_price)
security['ULTOSC'] = talib.ULTOSC(high_price, low_price, close_price)
security['WILLR'] = talib.WILLR(high_price, low_price, close_price)
security['NATR'] = talib.NATR(high_price, low_price, close_price)
security['RSI'] = talib.RSI(close_price)
security['EMA'] = talib.EMA(close_price)
security['SAREXT'] = talib.SAREXT(high_price, low_price)
security['RR'] = security[close_name] / security[close_name].shift(
1).fillna(1)
security['LOG_RR'] = np.log(security['RR'])
if volume_name:
security['MFI'] = talib.MFI(high_price, low_price, close_price,
volume)
security[volume_name] = np.log(security[volume_name])
security.drop([open_name, close_name, high_name, low_name], axis=1)
security = security.dropna().astype(np.float32)
return security
def ultosc(candles: np.ndarray, timeperiod1=7, timeperiod2=14, timeperiod3=28, sequential=False) -> Union[float, np.ndarray]:
"""
ULTOSC - Ultimate Oscillator
:param candles: np.ndarray
:param timeperiod1: int - default=7
:param timeperiod2: int - default=14
:param timeperiod3: int - default=28
:param sequential: bool - default=False
:return: float | np.ndarray
"""
if not sequential and len(candles) > 240:
candles = candles[-240:]
res = talib.ULTOSC(candles[:, 3], candles[:, 4], candles[:, 2], timeperiod1=timeperiod1, timeperiod2=timeperiod2, timeperiod3=timeperiod3)
if sequential:
return res
else:
return None if np.isnan(res[-1]) else res[-1]
def extract_features(data):
high = data['High']
low = data['Low']
close = data['Close']
volume = data['Volume']
open_ = data['Open']
data['ADX'] = ta.ADX(high, low, close, timeperiod=19)
data['CCI'] = ta.CCI(high, low, close, timeperiod=19)
data['CMO'] = ta.CMO(close, timeperiod=14)
data['MACD'], X, Y = ta.MACD(close,
fastperiod=10,
slowperiod=30,
signalperiod=9)
data['MFI'] = ta.MFI(high, low, close, volume, timeperiod=19)
data['MOM'] = ta.MOM(close, timeperiod=9)
data['ROCR'] = ta.ROCR(close, timeperiod=12)
data['RSI'] = ta.RSI(close, timeperiod=19)
data['STOCHSLOWK'], data['STOCHSLOWD'] = ta.STOCH(high,
low,
close,
fastk_period=5,
slowk_period=3,
slowk_matype=0,
slowd_period=3,
slowd_matype=0)
data['TRIX'] = ta.TRIX(close, timeperiod=30)
data['WILLR'] = ta.WILLR(high, low, close, timeperiod=14)
data['OBV'] = ta.OBV(close, volume)
data['TSF'] = ta.TSF(close, timeperiod=14)
data['NATR'] = ta.NATR(high, low, close) #, timeperiod=14)
data['ULTOSC'] = ta.ULTOSC(high, low, close)
data['AROONOSC'] = ta.AROONOSC(high, low, timeperiod=14)
data['BOP'] = ta.BOP(open_, high, low, close)
data['LINEARREG'] = ta.LINEARREG(close)
data['AP0'] = ta.APO(close, fastperiod=9, slowperiod=23, matype=1)
data['TEMA'] = ta.TRIMA(close, 29)
return data
def generate_tech_data(stock, open_name, close_name, high_name, low_name, max_time_window=10):
open_price = stock[open_name].values
close_price = stock[close_name].values
low_price = stock[low_name].values
high_price = stock[high_name].values
data = pd.DataFrame(stock)
data['MOM'] = talib.MOM(close_price, timeperiod=max_time_window)
# data['_SMA'] = talib.SMA(close_price)
data['HT_DCPERIOD'] = talib.HT_DCPERIOD(close_price)
data['HT_DCPHASE'] = talib.HT_DCPHASE(close_price)
data['sine'], data['leadsine'] = talib.HT_SINE(close_price)
data['inphase'], data['quadrature'] = talib.HT_PHASOR(close_price)
# data['_HT_TRENDMODE'] = talib.HT_TRENDMODE(close_price)
# data['_SAREXT'] = talib.SAREXT(high_price, low_price)
# data['_ADX'] = talib.ADX(high_price, low_price, close_price)
data['ADXR'] = talib.ADXR(high_price, low_price, close_price, timeperiod=max_time_window)
data['APO'] = talib.APO(close_price, fastperiod=max_time_window // 2, slowperiod=max_time_window)
data['AROON_UP'], _ = talib.AROON(high_price, low_price, timeperiod=max_time_window)
# data['_BOP'] = talib.BOP(open_price, high_price, low_price, close_price)
data['CCI'] = talib.CCI(high_price, low_price, close_price, timeperiod=max_time_window)
data['PLUS_DI'] = talib.PLUS_DI(high_price, low_price, close_price, timeperiod=max_time_window)
# data['_PLUS_DM'] = talib.PLUS_DM(high_price, low_price)
data['PPO'] = talib.PPO(close_price, fastperiod=max_time_window // 2, slowperiod=max_time_window)
data['macd'], data['macd_sig'], data['macd_hist'] = talib.MACD(close_price, fastperiod=max_time_window // 2, slowperiod=max_time_window, signalperiod=max_time_window // 2)
data['CMO'] = talib.CMO(close_price, timeperiod=max_time_window)
# data['ROC'] = talib.ROC(close_price)
data['ROCP'] = talib.ROCP(close_price, timeperiod=max_time_window)
# data['ROCR'] = talib.ROCR(close_price)
# data['slowk'], data['slowd'] = talib.STOCH(high_price, low_price, close_price)
data['fastk'], data['fastd'] = talib.STOCHF(high_price, low_price, close_price)
data['TRIX'] = talib.TRIX(close_price, timeperiod=max_time_window)
data['ULTOSC'] = talib.ULTOSC(high_price, low_price, close_price, timeperiod1=max_time_window // 2, timeperiod2=max_time_window, timeperiod3=max_time_window * 2)
data['WILLR'] = talib.WILLR(high_price, low_price, close_price, timeperiod=max_time_window)
data['NATR'] = talib.NATR(high_price, low_price, close_price, timeperiod=max_time_window)
# data['_TRANGE'] = talib.TRANGE(high_price, low_price, close_price)
data = data.drop([open_name, close_name, high_name, low_name], axis=1)
# data.columns=data.columns.map(lambda x:x[1:])
data = data.dropna().astype(np.float32)
return data
def create_Indicators(data):
dfs = data
df = dfs.copy()
df['ADX'] = talib.ADX(df.high, df.low, df.close, timeperiod=14)
df['ADXR'] = talib.ADXR(df.high, df.low, df.close, timeperiod=14)
df['APO'] = talib.APO(df.close, fastperiod=12, slowperiod=26, matype=0)
df['AROONOSC'] = talib.AROONOSC(df.high, df.low, timeperiod=14)
df['BOP'] = talib.BOP(
df.open, df.high, df.low,
df.close) # (Close price – Open price) / (High price – Low price)
df["CCI"] = talib.CCI(df.high, df.low, df.close, timeperiod=14)
df["CMO"] = talib.CMO(df.close, timeperiod=14)
df["DX"] = talib.DX(df.high, df.low, df.close, timeperiod=14)
#df["MFI"] = talib.MFI(df.high, df.low, df.close, df.volume, timeperiod=14).pct_change()
df["MINUS_DI"] = talib.MINUS_DI(df.high, df.low, df.close, timeperiod=14)
df["MINUS_DM"] = talib.MINUS_DM(df.high, df.low, timeperiod=14)
df["MOM"] = talib.MOM(df.close, timeperiod=10)
df["PLUS_DI"] = talib.PLUS_DI(df.high, df.low, df.close, timeperiod=14)
df["PLUS_DM"] = talib.PLUS_DM(df.high, df.low, timeperiod=14)
#X_corr["PPO"] = PPO(df.close, fastperiod=12, sdf.lowperiod=26, matype=0)
df["ROC"] = talib.ROC(df.close, timeperiod=10)
df["ROCP"] = talib.ROCP(df.close, timeperiod=10)
df["ROCR"] = talib.ROCR(df.close, timeperiod=10)
df["ROCR100"] = talib.ROCR100(df.close, timeperiod=10)
df["RSI"] = talib.RSI(df.close, timeperiod=14)
#sdf.lowk, sdf.lowd = STOCH(df.high, df.low, df.close, fastk_period=5, sdf.lowk_period=3, sdf.lowk_matype=0, sdf.lowd_period=3, sdf.lowd_matype=0)
df["TRIX"] = talib.TRIX(df.close, timeperiod=30)
df["ULTOSC"] = talib.ULTOSC(df.high,
df.low,
df.close,
timeperiod1=7,
timeperiod2=14,
timeperiod3=28)
df["WILLR"] = talib.WILLR(df.high, df.low, df.close, timeperiod=14)
return df
def TKE(dataframe, *, length=14, emaperiod=5):
"""
Source: https://www.tradingview.com/script/Pcbvo0zG/
Author: Dr Yasar ERDINC
The calculation is simple:
TKE=(RSI+STOCHASTIC+ULTIMATE OSCILLATOR+MFI+WIILIAMS %R+MOMENTUM+CCI)/7
Buy signal: when TKE crosses above 20 value
Oversold region: under 20 value
Overbought region: over 80 value
Another usage of TKE is with its EMA ,
the default value is defined as 5 bars of EMA of the TKE line,
Go long: when TKE crosses above EMALine
Go short: when TKE crosses below EMALine
Usage:
`dataframe['TKE'], dataframe['TKEema'] = TKE1(dataframe)`
"""
import talib.abstract as ta
df = dataframe.copy()
# TKE=(RSI+STOCHASTIC+ULTIMATE OSCILLATOR+MFI+WIILIAMS %R+MOMENTUM+CCI)/7
df["rsi"] = ta.RSI(df, timeperiod=length)
df["stoch"] = (100 *
(df["close"] - df["low"].rolling(window=length).min()) /
(df["high"].rolling(window=length).max() -
df["low"].rolling(window=length).min()))
df["ultosc"] = ta.ULTOSC(df, timeperiod1=7, timeperiod2=14, timeperiod3=28)
df["mfi"] = ta.MFI(df, timeperiod=length)
df["willr"] = ta.WILLR(df, timeperiod=length)
df["mom"] = ta.ROCR100(df, timeperiod=length)
df["cci"] = ta.CCI(df, timeperiod=length)
df["TKE"] = df[["rsi", "stoch", "ultosc", "mfi", "willr", "mom",
"cci"]].mean(axis="columns")
df["TKEema"] = ta.EMA(df["TKE"], timeperiod=emaperiod)
return df["TKE"], df["TKEema"]
def add_ULTOSC(self,
timeperiod=14,
timeperiod2=14,
timeperiod3=28,
type='line',
color='secondary',
**kwargs):
"""Ultimate Oscillator."""
if not (self.has_high and self.has_low and self.has_close):
raise Exception()
utils.kwargs_check(kwargs, VALID_TA_KWARGS)
if 'kind' in kwargs:
type = kwargs['kind']
name = 'ULTOSC({})'.format(str(timeperiod), str(timeperiod2),
str(timeperiod3))
self.sec[name] = dict(type=type, color=color)
self.ind[name] = talib.ULTOSC(self.df[self.hi].values,
self.df[self.lo].values,
self.df[self.cl].values, timeperiod,
timeperiod2, timeperiod3)
def Feature_Extraction_Momentum(df):
df = SOK(df)
df = SOD(df)
df = SOJ(df)
df = MACD(df)
df = MACD_Signal(df)
df = MACD_Hist(df)
df['RSI'] = tas.RSI(df['close'].values)
df['ROC'] = tas.ROC(df['close'].values)
df['WillR'] = tas.WILLR(df['high'].values, df['low'].values,
df['close'].values)
df['CCI'] = tas.CCI(df['high'].values, df['low'].values,
df['close'].values)
df['TSI'] = tsi(df['close'], window_slow=25, window_fast=13)
df['ADX'] = tas.ADX(df['high'].values, df['low'].values,
df['close'].values)
df['MFI'] = tas.MFI(df['high'], df['low'], df['close'], df['volume'])
df['MOM'] = tas.MOM(df['close'])
df['TRIX'] = tas.TRIX(df['close'])
df['ULTOSC'] = tas.ULTOSC(df['high'], df['low'], df['close'])
df = df.fillna(0)
return df
def UltOsc(self, df, t1=7, t2=14, t3=28):
""" Ultimate Oscillator
Uses weighted sums of three oscillators, designed to capture
momentum across three different timeframes, each of which uses
a different time period
Args:
high, low, close: HLC of instrument
t1, t2, t3: various time periods in the calculation,
default: 7,14,28
feature_dict: Dictionary of added features
Return:
UO signal
feature_dict
"""
t1t = 'UltOsc_t1' + str(t1)
t2t = '_t2' + str(t2)
t3t = '_t3' + str(t3)
col_name = t1t + t2t + t3t
current_feature['Latest'] = col_name
feature_dict[col_name] = 'Keep'
df[col_name] = ta.ULTOSC(df.High, df.Low, df.Close, t1, t2, t3)
return df
def preprocess_data(self):
"""calculate returns and percentiles, then removes missing values"""
self.data['returns'] = self.data.Close.pct_change()
self.data['ret_2'] = self.data.Close.pct_change(2)
self.data['ret_5'] = self.data.Close.pct_change(5)
self.data['ret_10'] = self.data.Close.pct_change(10)
self.data['ret_21'] = self.data.Close.pct_change(21)
self.data['rsi'] = talib.STOCHRSI(self.data.Close)[1]
self.data['macd'] = talib.MACD(self.data.Close)[1]
self.data['atr'] = talib.ATR(self.data.High, self.data.Low,
self.data.Close)
slowk, slowd = talib.STOCH(self.data.High, self.data.Low,
self.data.Close)
self.data['stoch'] = slowd - slowk
self.data['atr'] = talib.ATR(self.data.High, self.data.Low,
self.data.Close)
self.data['ultosc'] = talib.ULTOSC(self.data.High, self.data.Low,
self.data.Close)
self.data = (self.data.replace(
(np.inf, -np.inf), np.nan).drop([
'High', 'Low', 'Close', 'Volume', 'Symbol', 'Volume ETH',
'Volume BTC', 'Unix Timestamp'
],
axis=1).dropna())
r = self.data.returns.copy()
if self.normalize:
self.data = pd.DataFrame(scale(self.data),
columns=self.data.columns,
index=self.data.index)
features = self.data.columns.drop('returns')
self.data['returns'] = r # don't scale returns
self.data = self.data.loc[:, ['returns'] + list(features)]
log.info(self.data.info())
def extract_by_type(feature_type,
open_prices=None,
close_prices=None,
high_prices=None,
low_prices=None):
if feature_type == 'ROCP':
rocp1 = talib.ROCP(close_prices, timeperiod=1)
rocp2 = talib.ROCP(close_prices, timeperiod=2)
feature.append(rocp1)
feature.append(rocp2)
if feature_type == 'OROCP':
orocp = talib.ROCP(open_prices, timeperiod=1)
feature.append(orocp)
if feature_type == 'HROCP':
hrocp = talib.ROCP(high_prices, timeperiod=1)
feature.append(hrocp)
if feature_type == 'LROCP':
lrocp = talib.ROCP(low_prices, timeperiod=1)
feature.append(lrocp)
if feature_type == 'MACD':
macd, signal, hist = talib.MACD(close_prices,
fastperiod=12,
slowperiod=26,
signalperiod=9)
norm_macd = numpy.nan_to_num(macd) / math.sqrt(
numpy.var(numpy.nan_to_num(macd)))
norm_signal = numpy.nan_to_num(signal) / math.sqrt(
numpy.var(numpy.nan_to_num(signal)))
norm_hist = numpy.nan_to_num(hist) / math.sqrt(
numpy.var(numpy.nan_to_num(hist)))
macdrocp = talib.ROCP(norm_macd + numpy.max(norm_macd) -
numpy.min(norm_macd),
timeperiod=1)
signalrocp = talib.ROCP(norm_signal + numpy.max(norm_signal) -
numpy.min(norm_signal),
timeperiod=1)
histrocp = talib.ROCP(norm_hist + numpy.max(norm_hist) -
numpy.min(norm_hist),
timeperiod=1)
# feature.append(macd / 100.0)
# feature.append(signal / 100.0)
# feature.append(hist / 100.0)
feature.append(norm_macd)
feature.append(norm_signal)
feature.append(norm_hist)
feature.append(macdrocp)
feature.append(signalrocp)
feature.append(histrocp)
if feature_type == 'RSI':
rsi6 = talib.RSI(close_prices, timeperiod=6)
rsi12 = talib.RSI(close_prices, timeperiod=12)
rsi24 = talib.RSI(close_prices, timeperiod=24)
rsi6rocp = talib.ROCP(rsi6 + 100., timeperiod=1)
rsi12rocp = talib.ROCP(rsi12 + 100., timeperiod=1)
rsi24rocp = talib.ROCP(rsi24 + 100., timeperiod=1)
feature.append(rsi6 / 100.0 - 0.5)
feature.append(rsi12 / 100.0 - 0.5)
feature.append(rsi24 / 100.0 - 0.5)
# feature.append(numpy.maximum(rsi6 / 100.0 - 0.8, 0))
# feature.append(numpy.maximum(rsi12 / 100.0 - 0.8, 0))
# feature.append(numpy.maximum(rsi24 / 100.0 - 0.8, 0))
# feature.append(numpy.minimum(rsi6 / 100.0 - 0.2, 0))
# feature.append(numpy.minimum(rsi6 / 100.0 - 0.2, 0))
# feature.append(numpy.minimum(rsi6 / 100.0 - 0.2, 0))
# feature.append(numpy.maximum(numpy.minimum(rsi6 / 100.0 - 0.5, 0.3), -0.3))
# feature.append(numpy.maximum(numpy.minimum(rsi6 / 100.0 - 0.5, 0.3), -0.3))
# feature.append(numpy.maximum(numpy.minimum(rsi6 / 100.0 - 0.5, 0.3), -0.3))
feature.append(rsi6rocp)
feature.append(rsi12rocp)
feature.append(rsi24rocp)
if feature_type == 'UO':
ult_osc = talib.ULTOSC(high_prices,
low_prices,
close_prices,
timeperiod1=7,
timeperiod2=14,
timeperiod3=28)
feature.append(ult_osc / 100.0 - 0.5)
if feature_type == 'BOLL':
upperband, middleband, lowerband = talib.BBANDS(close_prices,
timeperiod=5,
nbdevup=2,
nbdevdn=2,
matype=0)
feature.append((upperband - close_prices) / close_prices)
feature.append((middleband - close_prices) / close_prices)
feature.append((lowerband - close_prices) / close_prices)
if feature_type == 'MA':
ma5 = talib.MA(close_prices, timeperiod=5)
ma10 = talib.MA(close_prices, timeperiod=10)
ma20 = talib.MA(close_prices, timeperiod=20)
ma25 = talib.MA(close_prices, timeperiod=25)
ma30 = talib.MA(close_prices, timeperiod=30)
ma40 = talib.MA(close_prices, timeperiod=40)
ma50 = talib.MA(close_prices, timeperiod=50)
ma60 = talib.MA(close_prices, timeperiod=60)
#ma360 = talib.MA(close_prices, timeperiod=70)
#ma720 = talib.MA(close_prices, timeperiod=720)
ma5rocp = talib.ROCP(ma5, timeperiod=1)
ma10rocp = talib.ROCP(ma10, timeperiod=1)
ma20rocp = talib.ROCP(ma20, timeperiod=1)
ma25rocp = talib.ROCP(ma25, timeperiod=1)
ma30rocp = talib.ROCP(ma30, timeperiod=1)
ma40rocp = talib.ROCP(ma40, timeperiod=1)
ma50rocp = talib.ROCP(ma50, timeperiod=1)
ma60rocp = talib.ROCP(ma60, timeperiod=1)
#ma360rocp = talib.ROCP(ma360, timeperiod=1)
#ma720rocp = talib.ROCP(ma720, timeperiod=1)
feature.append(ma5rocp)
feature.append(ma10rocp)
feature.append(ma20rocp)
feature.append(ma25rocp)
feature.append(ma30rocp)
feature.append(ma40rocp)
feature.append(ma50rocp)
feature.append(ma60rocp)
#feature.append(ma360rocp)
#feature.append(ma720rocp)
feature.append((ma5 - close_prices) / close_prices)
feature.append((ma10 - close_prices) / close_prices)
feature.append((ma20 - close_prices) / close_prices)
feature.append((ma25 - close_prices) / close_prices)
feature.append((ma30 - close_prices) / close_prices)
feature.append((ma40 - close_prices) / close_prices)
feature.append((ma50 - close_prices) / close_prices)
feature.append((ma60 - close_prices) / close_prices)
#feature.append((ma360 - close_prices) / close_prices)
#feature.append((ma720 - close_prices) / close_prices)
if feature_type == 'STOCH':
slow_stoch_k, slow_stoch_d = talib.STOCH(high_prices,
low_prices,
close_prices,
fastk_period=5,
slowk_period=3,
slowk_matype=0,
slowd_period=3,
slowd_matype=0)
fast_stoch_k, fast_stoch_d = talib.STOCHF(high_prices,
low_prices,
close_prices,
fastk_period=5,
fastd_period=3,
fastd_matype=0)
fast_rsi_k, fast_rsi_d = talib.STOCHRSI(close_prices,
timeperiod=14,
fastk_period=5,
fastd_period=3,
fastd_matype=0)
feature.append(slow_stoch_k / 100.0 - 0.5)
feature.append(slow_stoch_d / 100.0 - 0.5)
feature.append(fast_stoch_k / 100.0 - 0.5)
feature.append(fast_stoch_d / 100.0 - 0.5)
feature.append(fast_rsi_k / 100.0 - 0.5)
feature.append(fast_rsi_d / 100.0 - 0.5)
if feature_type == 'AO':
median_price = (high_prices + low_prices) / 2
ao = talib.SMA(median_price, 5) - talib.SMA(median_price, 34)
feature.append(ao)
if feature_type == 'ROC':
roc5 = talib.ROC(close_prices, timeperiod=5)
roc10 = talib.ROC(close_prices, timeperiod=10)
roc20 = talib.ROC(close_prices, timeperiod=20)
roc25 = talib.ROC(close_prices, timeperiod=25)
feature.append(roc5)
feature.append(roc10)
feature.append(roc20)
feature.append(roc25)
if feature_type == 'WILLR':
willr = talib.WILLR(high_prices,
low_prices,
close_prices,
timeperiod=14)
feature.append(willr / 100.0 - 0.5)
return feature
def generate_tech_data(stock,
open_name,
close_name,
high_name,
low_name,
max_time_window=10):
open_price = stock[open_name].values
close_price = stock[close_name].values
low_price = stock[low_name].values
high_price = stock[high_name].values
data = stock.copy()
data['MOM'] = talib.MOM(close_price, timeperiod=max_time_window)
data['HT_DCPERIOD'] = talib.HT_DCPERIOD(close_price)
data['HT_DCPHASE'] = talib.HT_DCPHASE(close_price)
data['sine'], data['leadsine'] = talib.HT_SINE(close_price)
data['inphase'], data['quadrature'] = talib.HT_PHASOR(close_price)
data['ADXR'] = talib.ADXR(high_price,
low_price,
close_price,
timeperiod=max_time_window)
data['APO'] = talib.APO(close_price,
fastperiod=max_time_window // 2,
slowperiod=max_time_window)
data['AROON_UP'], _ = talib.AROON(high_price,
low_price,
timeperiod=max_time_window)
data['CCI'] = talib.CCI(high_price,
low_price,
close_price,
timeperiod=max_time_window)
data['PLUS_DI'] = talib.PLUS_DI(high_price,
low_price,
close_price,
timeperiod=max_time_window)
data['PPO'] = talib.PPO(close_price,
fastperiod=max_time_window // 2,
slowperiod=max_time_window)
data['macd'], data['macd_sig'], data['macd_hist'] = talib.MACD(
close_price,
fastperiod=max_time_window // 2,
slowperiod=max_time_window,
signalperiod=max_time_window // 2)
data['CMO'] = talib.CMO(close_price, timeperiod=max_time_window)
data['ROCP'] = talib.ROCP(close_price, timeperiod=max_time_window)
data['fastk'], data['fastd'] = talib.STOCHF(high_price, low_price,
close_price)
data['TRIX'] = talib.TRIX(close_price, timeperiod=max_time_window)
data['ULTOSC'] = talib.ULTOSC(high_price,
low_price,
close_price,
timeperiod1=max_time_window // 2,
timeperiod2=max_time_window,
timeperiod3=max_time_window * 2)
data['WILLR'] = talib.WILLR(high_price,
low_price,
close_price,
timeperiod=max_time_window)
data['NATR'] = talib.NATR(high_price,
low_price,
close_price,
timeperiod=max_time_window)
data = data.drop([open_name, close_name, high_name, low_name], axis=1)
data = data.dropna().astype(np.float32)
return data
def get_factors(index,
Open,
Close,
High,
Low,
Volume,
rolling=26,
drop=False,
normalization=True):
tmp = pd.DataFrame()
tmp['tradeTime'] = index
# 累积/派发线（Accumulation / Distribution Line，该指标将每日的成交量通过价格加权累计，
# 用以计算成交量的动量。属于趋势型因子
tmp['AD'] = talib.AD(High, Low, Close, Volume)
# 佳庆指标（Chaikin Oscillator），该指标基于AD曲线的指数移动均线而计算得到。属于趋势型因子
tmp['ADOSC'] = talib.ADOSC(High,
Low,
Close,
Volume,
fastperiod=3,
slowperiod=10)
# 平均动向指数，DMI因子的构成部分。属于趋势型因子
tmp['ADX'] = talib.ADX(High, Low, Close, timeperiod=14)
# 相对平均动向指数，DMI因子的构成部分。属于趋势型因子
tmp['ADXR'] = talib.ADXR(High, Low, Close, timeperiod=14)
# 绝对价格振荡指数
tmp['APO'] = talib.APO(Close, fastperiod=12, slowperiod=26)
# Aroon通过计算自价格达到近期最高值和最低值以来所经过的期间数，帮助投资者预测证券价格从趋势到区域区域或反转的变化，
# Aroon指标分为Aroon、AroonUp和AroonDown3个具体指标。属于趋势型因子
tmp['AROONDown'], tmp['AROONUp'] = talib.AROON(High, Low, timeperiod=14)
tmp['AROONOSC'] = talib.AROONOSC(High, Low, timeperiod=14)
# 均幅指标（Average TRUE Ranger），取一定时间周期内的股价波动幅度的移动平均值，
# 是显示市场变化率的指标，主要用于研判买卖时机。属于超买超卖型因子。
tmp['ATR14'] = talib.ATR(High, Low, Close, timeperiod=14)
tmp['ATR6'] = talib.ATR(High, Low, Close, timeperiod=6)
# 布林带
tmp['Boll_Up'], tmp['Boll_Mid'], tmp['Boll_Down'] = talib.BBANDS(
Close, timeperiod=20, nbdevup=2, nbdevdn=2, matype=0)
# 均势指标
tmp['BOP'] = talib.BOP(Open, High, Low, Close)
# 5日顺势指标（Commodity Channel Index），专门测量股价是否已超出常态分布范围。属于超买超卖型因子。
tmp['CCI5'] = talib.CCI(High, Low, Close, timeperiod=5)
tmp['CCI10'] = talib.CCI(High, Low, Close, timeperiod=10)
tmp['CCI20'] = talib.CCI(High, Low, Close, timeperiod=20)
tmp['CCI88'] = talib.CCI(High, Low, Close, timeperiod=88)
# 钱德动量摆动指标（Chande Momentum Osciliator），与其他动量指标摆动指标如相对强弱指标（RSI）和随机指标（KDJ）不同，
# 钱德动量指标在计算公式的分子中采用上涨日和下跌日的数据。属于超买超卖型因子
tmp['CMO_Close'] = talib.CMO(Close, timeperiod=14)
tmp['CMO_Open'] = talib.CMO(Close, timeperiod=14)
# DEMA双指数移动平均线
tmp['DEMA6'] = talib.DEMA(Close, timeperiod=6)
tmp['DEMA12'] = talib.DEMA(Close, timeperiod=12)
tmp['DEMA26'] = talib.DEMA(Close, timeperiod=26)
# DX 动向指数
tmp['DX'] = talib.DX(High, Low, Close, timeperiod=14)
# EMA 指数移动平均线
tmp['EMA6'] = talib.EMA(Close, timeperiod=6)
tmp['EMA12'] = talib.EMA(Close, timeperiod=12)
tmp['EMA26'] = talib.EMA(Close, timeperiod=26)
# KAMA 适应性移动平均线
tmp['KAMA'] = talib.KAMA(Close, timeperiod=30)
# MACD
tmp['MACD_DIF'], tmp['MACD_DEA'], tmp['MACD_bar'] = talib.MACD(
Close, fastperiod=12, slowperiod=24, signalperiod=9)
# 中位数价格 不知道是什么意思
tmp['MEDPRICE'] = talib.MEDPRICE(High, Low)
# 负向指标 负向运动
tmp['MiNUS_DI'] = talib.MINUS_DI(High, Low, Close, timeperiod=14)
tmp['MiNUS_DM'] = talib.MINUS_DM(High, Low, timeperiod=14)
# 动量指标（Momentom Index），动量指数以分析股价波动的速度为目的，研究股价在波动过程中各种加速，
# 减速，惯性作用以及股价由静到动或由动转静的现象。属于趋势型因子
tmp['MOM'] = talib.MOM(Close, timeperiod=10)
# 归一化平均值范围
tmp['NATR'] = talib.NATR(High, Low, Close, timeperiod=14)
# OBV 能量潮指标（On Balance Volume，OBV），以股市的成交量变化来衡量股市的推动力，
# 从而研判股价的走势。属于成交量型因子
tmp['OBV'] = talib.OBV(Close, Volume)
# PLUS_DI 更向指示器
tmp['PLUS_DI'] = talib.PLUS_DI(High, Low, Close, timeperiod=14)
tmp['PLUS_DM'] = talib.PLUS_DM(High, Low, timeperiod=14)
# PPO 价格振荡百分比
tmp['PPO'] = talib.PPO(Close, fastperiod=6, slowperiod=26, matype=0)
# ROC 6日变动速率（Price Rate of Change），以当日的收盘价和N天前的收盘价比较，
# 通过计算股价某一段时间内收盘价变动的比例，应用价格的移动比较来测量价位动量。属于超买超卖型因子。
tmp['ROC6'] = talib.ROC(Close, timeperiod=6)
tmp['ROC20'] = talib.ROC(Close, timeperiod=20)
# 12日量变动速率指标（Volume Rate of Change），以今天的成交量和N天前的成交量比较，
# 通过计算某一段时间内成交量变动的幅度，应用成交量的移动比较来测量成交量运动趋向，
# 达到事先探测成交量供需的强弱，进而分析成交量的发展趋势及其将来是否有转势的意愿，
# 属于成交量的反趋向指标。属于成交量型因子
tmp['VROC6'] = talib.ROC(Volume, timeperiod=6)
tmp['VROC20'] = talib.ROC(Volume, timeperiod=20)
# ROC 6日变动速率（Price Rate of Change），以当日的收盘价和N天前的收盘价比较，
# 通过计算股价某一段时间内收盘价变动的比例，应用价格的移动比较来测量价位动量。属于超买超卖型因子。
tmp['ROCP6'] = talib.ROCP(Close, timeperiod=6)
tmp['ROCP20'] = talib.ROCP(Close, timeperiod=20)
# 12日量变动速率指标（Volume Rate of Change），以今天的成交量和N天前的成交量比较，
# 通过计算某一段时间内成交量变动的幅度，应用成交量的移动比较来测量成交量运动趋向，
# 达到事先探测成交量供需的强弱，进而分析成交量的发展趋势及其将来是否有转势的意愿，
# 属于成交量的反趋向指标。属于成交量型因子
tmp['VROCP6'] = talib.ROCP(Volume, timeperiod=6)
tmp['VROCP20'] = talib.ROCP(Volume, timeperiod=20)
# RSI
tmp['RSI'] = talib.RSI(Close, timeperiod=14)
# SAR 抛物线转向
tmp['SAR'] = talib.SAR(High, Low, acceleration=0.02, maximum=0.2)
# TEMA
tmp['TEMA6'] = talib.TEMA(Close, timeperiod=6)
tmp['TEMA12'] = talib.TEMA(Close, timeperiod=12)
tmp['TEMA26'] = talib.TEMA(Close, timeperiod=26)
# TRANGE 真实范围
tmp['TRANGE'] = talib.TRANGE(High, Low, Close)
# TYPPRICE 典型价格
tmp['TYPPRICE'] = talib.TYPPRICE(High, Low, Close)
# TSF 时间序列预测
tmp['TSF'] = talib.TSF(Close, timeperiod=14)
# ULTOSC 极限振子
tmp['ULTOSC'] = talib.ULTOSC(High,
Low,
Close,
timeperiod1=7,
timeperiod2=14,
timeperiod3=28)
# 威廉指标
tmp['WILLR'] = talib.WILLR(High, Low, Close, timeperiod=14)
# 标准化
if normalization:
factors_list = tmp.columns.tolist()[1:]
if rolling >= 26:
for i in factors_list:
tmp[i] = (tmp[i] - tmp[i].rolling(window=rolling, center=False).mean()) \
/ tmp[i].rolling(window=rolling, center=False).std()
elif rolling < 26 & rolling > 0:
print('Recommended rolling range greater than 26')
elif rolling <= 0:
for i in factors_list:
tmp[i] = (tmp[i] - tmp[i].mean()) / tmp[i].std()
if drop:
tmp.dropna(inplace=True)
tmp.set_index('tradeTime', inplace=True)
return tmp
def main():
ohlcv = api_ohlcv('20191017')
open, high, low, close, volume, timestamp = [], [], [], [], [], []
for i in ohlcv:
open.append(int(i[0]))
high.append(int(i[1]))
low.append(int(i[2]))
close.append(int(i[3]))
volume.append(float(i[4]))
time_str = str(i[5])
timestamp.append(
datetime.fromtimestamp(int(
time_str[:10])).strftime('%Y/%m/%d %H:%M:%M'))
date_time_index = pd.to_datetime(
timestamp) # convert to DateTimeIndex type
df = pd.DataFrame(
{
'open': open,
'high': high,
'low': low,
'close': close,
'volume': volume
},
index=date_time_index)
# df.index += pd.offsets.Hour(9) # adjustment for JST if required
print(df.shape)
print(df.columns)
# pct_change
f = lambda x: 1 if x > 0.0001 else -1 if x < -0.0001 else 0 if -0.0001 <= x <= 0.0001 else np.nan
y = df.rename(columns={
'close': 'y'
}).loc[:, 'y'].pct_change(1).shift(-1).fillna(0)
X = df.copy()
y_ = pd.DataFrame(y.map(f), columns=['y'])
y = df.rename(columns={'close': 'y'}).loc[:, 'y'].pct_change(1).fillna(0)
df_ = pd.concat([X, y_], axis=1)
# check the shape
print(
'----------------------------------------------------------------------------------------'
)
print('X shape: (%i,%i)' % X.shape)
print('y shape: (%i,%i)' % y_.shape)
print(
'----------------------------------------------------------------------------------------'
)
print(y_.groupby('y').size())
print('y=1 up, y=0 stay, y=-1 down')
print(
'----------------------------------------------------------------------------------------'
)
# feature calculation
open = pd.Series(df['open'])
high = pd.Series(df['high'])
low = pd.Series(df['low'])
close = pd.Series(df['close'])
volume = pd.Series(df['volume'])
# pct_change for new column
X['diff'] = y
# Exponential Moving Average
ema = talib.EMA(close, timeperiod=3)
ema = ema.fillna(ema.mean())
# Momentum
momentum = talib.MOM(close, timeperiod=5)
momentum = momentum.fillna(momentum.mean())
# RSI
rsi = talib.RSI(close, timeperiod=14)
rsi = rsi.fillna(rsi.mean())
# ADX
adx = talib.ADX(high, low, close, timeperiod=14)
adx = adx.fillna(adx.mean())
# ADX change
adx_change = adx.pct_change(1).shift(-1)
adx_change = adx_change.fillna(adx_change.mean())
# AD
ad = talib.AD(high, low, close, volume)
ad = ad.fillna(ad.mean())
X_ = pd.concat([X, ema, momentum, rsi, adx_change, ad],
axis=1).drop(['open', 'high', 'low', 'close'], axis=1)
X_.columns = ['volume', 'diff', 'ema', 'momentum', 'rsi', 'adx', 'ad']
X_.join(y_).head(10)
# default parameter models
X_train, X_test, y_train, y_test = train_test_split(X_,
y_,
test_size=0.33,
random_state=42)
print('X_train shape: {}'.format(X_train.shape))
print('X_test shape: {}'.format(X_test.shape))
print('y_train shape: {}'.format(y_train.shape))
print('y_test shape: {}'.format(y_test.shape))
pipe_knn = Pipeline([('scl', StandardScaler()),
('est', KNeighborsClassifier(n_neighbors=3))])
pipe_logistic = Pipeline([('scl', StandardScaler()),
('est',
LogisticRegression(solver='lbfgs',
multi_class='multinomial',
random_state=39))])
pipe_rf = Pipeline([('scl', StandardScaler()),
('est', RandomForestClassifier(random_state=39))])
pipe_gb = Pipeline([('scl', StandardScaler()),
('est', GradientBoostingClassifier(random_state=39))])
pipe_names = ['KNN', 'Logistic', 'RandomForest', 'GradientBoosting']
pipe_lines = [pipe_knn, pipe_logistic, pipe_rf, pipe_gb]
for (i, pipe) in enumerate(pipe_lines):
pipe.fit(X_train, y_train.values.ravel())
print(pipe)
print('%s: %.3f' %
(pipe_names[i] + ' Train Accuracy',
accuracy_score(y_train.values.ravel(), pipe.predict(X_train))))
print('%s: %.3f' %
(pipe_names[i] + ' Test Accuracy',
accuracy_score(y_test.values.ravel(), pipe.predict(X_test))))
print('%s: %.3f' % (pipe_names[i] + ' Train F1 Score',
f1_score(y_train.values.ravel(),
pipe.predict(X_train),
average='micro')))
print('%s: %.3f' % (pipe_names[i] + ' Test F1 Score',
f1_score(y_test.values.ravel(),
pipe.predict(X_test),
average='micro')))
for (i, pipe) in enumerate(pipe_lines):
predict = pipe.predict(X_test)
cm = confusion_matrix(y_test.values.ravel(),
predict,
labels=[-1, 0, 1])
print('{} Confusion Matrix'.format(pipe_names[i]))
print(cm)
## Overlap Studies Functions
# DEMA - Double Exponential Moving Average
dema = talib.DEMA(close, timeperiod=3)
dema = dema.fillna(dema.mean())
print('DEMA - Double Exponential Moving Average shape: {}'.format(
dema.shape))
# EMA - Exponential Moving Average
ema = talib.EMA(close, timeperiod=3)
ema = ema.fillna(ema.mean())
print('EMA - Exponential Moving Average shape: {}'.format(ema.shape))
# HT_TRENDLINE - Hilbert Transform - Instantaneous Trendline
hilbert = talib.HT_TRENDLINE(close)
hilbert = hilbert.fillna(hilbert.mean())
print(
'HT_TRENDLINE - Hilbert Transform - Instantaneous Trendline shape: {}'.
format(hilbert.shape))
# KAMA - Kaufman Adaptive Moving Average
kama = talib.KAMA(close, timeperiod=3)
kama = kama.fillna(kama.mean())
print('KAMA - Kaufman Adaptive Moving Average shape: {}'.format(
kama.shape))
# MA - Moving average
ma = talib.MA(close, timeperiod=3, matype=0)
ma = ma.fillna(ma.mean())
print('MA - Moving average shape: {}'.format(kama.shape))
# MIDPOINT - MidPoint over period
midpoint = talib.MIDPOINT(close, timeperiod=7)
midpoint = midpoint.fillna(midpoint.mean())
print('MIDPOINT - MidPoint over period shape: {}'.format(midpoint.shape))
# MIDPRICE - Midpoint Price over period
midprice = talib.MIDPRICE(high, low, timeperiod=7)
midprice = midprice.fillna(midprice.mean())
print('MIDPRICE - Midpoint Price over period shape: {}'.format(
midprice.shape))
# SAR - Parabolic SAR
sar = talib.SAR(high, low, acceleration=0, maximum=0)
sar = sar.fillna(sar.mean())
print('SAR - Parabolic SAR shape: {}'.format(sar.shape))
# SAREXT - Parabolic SAR - Extended
sarext = talib.SAREXT(high,
low,
startvalue=0,
offsetonreverse=0,
accelerationinitlong=0,
accelerationlong=0,
accelerationmaxlong=0,
accelerationinitshort=0,
accelerationshort=0,
accelerationmaxshort=0)
sarext = sarext.fillna(sarext.mean())
print('SAREXT - Parabolic SAR - Extended shape: {}'.format(sarext.shape))
# SMA - Simple Moving Average
sma = talib.SMA(close, timeperiod=3)
sma = sma.fillna(sma.mean())
print('SMA - Simple Moving Average shape: {}'.format(sma.shape))
# T3 - Triple Exponential Moving Average (T3)
t3 = talib.T3(close, timeperiod=5, vfactor=0)
t3 = t3.fillna(t3.mean())
print('T3 - Triple Exponential Moving Average shape: {}'.format(t3.shape))
# TEMA - Triple Exponential Moving Average
tema = talib.TEMA(close, timeperiod=3)
tema = tema.fillna(tema.mean())
print('TEMA - Triple Exponential Moving Average shape: {}'.format(
tema.shape))
# TRIMA - Triangular Moving Average
trima = talib.TRIMA(close, timeperiod=3)
trima = trima.fillna(trima.mean())
print('TRIMA - Triangular Moving Average shape: {}'.format(trima.shape))
# WMA - Weighted Moving Average
wma = talib.WMA(close, timeperiod=3)
wma = wma.fillna(wma.mean())
print('WMA - Weighted Moving Average shape: {}'.format(wma.shape))
## Momentum Indicator Functions
# ADX - Average Directional Movement Index
adx = talib.ADX(high, low, close, timeperiod=14)
adx = adx.fillna(adx.mean())
print('ADX - Average Directional Movement Index shape: {}'.format(
adx.shape))
# ADXR - Average Directional Movement Index Rating
adxr = talib.ADXR(high, low, close, timeperiod=7)
adxr = adxr.fillna(adxr.mean())
print('ADXR - Average Directional Movement Index Rating shape: {}'.format(
adxr.shape))
# APO - Absolute Price Oscillator
apo = talib.APO(close, fastperiod=12, slowperiod=26, matype=0)
apo = apo.fillna(apo.mean())
print('APO - Absolute Price Oscillator shape: {}'.format(apo.shape))
# AROONOSC - Aroon Oscillator
aroon = talib.AROONOSC(high, low, timeperiod=14)
aroon = aroon.fillna(aroon.mean())
print('AROONOSC - Aroon Oscillator shape: {}'.format(apo.shape))
# BOP - Balance Of Power
bop = talib.BOP(open, high, low, close)
bop = bop.fillna(bop.mean())
print('BOP - Balance Of Power shape: {}'.format(apo.shape))
# CCI - Commodity Channel Index
cci = talib.CCI(high, low, close, timeperiod=7)
cci = cci.fillna(cci.mean())
print('CCI - Commodity Channel Index shape: {}'.format(cci.shape))
# CMO - Chande Momentum Oscillator
cmo = talib.CMO(close, timeperiod=7)
cmo = cmo.fillna(cmo.mean())
print('CMO - Chande Momentum Oscillator shape: {}'.format(cmo.shape))
# DX - Directional Movement Index
dx = talib.DX(high, low, close, timeperiod=7)
dx = dx.fillna(dx.mean())
print('DX - Directional Movement Index shape: {}'.format(dx.shape))
# MFI - Money Flow Index
mfi = talib.MFI(high, low, close, volume, timeperiod=7)
mfi = mfi.fillna(mfi.mean())
print('MFI - Money Flow Index shape: {}'.format(mfi.shape))
# MINUS_DI - Minus Directional Indicator
minusdi = talib.MINUS_DI(high, low, close, timeperiod=14)
minusdi = minusdi.fillna(minusdi.mean())
print('MINUS_DI - Minus Directional Indicator shape: {}'.format(
minusdi.shape))
# MINUS_DM - Minus Directional Movement
minusdm = talib.MINUS_DM(high, low, timeperiod=14)
minusdm = minusdm.fillna(minusdm.mean())
print('MINUS_DM - Minus Directional Movement shape: {}'.format(
minusdm.shape))
# MOM - Momentum
mom = talib.MOM(close, timeperiod=5)
mom = mom.fillna(mom.mean())
print('MOM - Momentum shape: {}'.format(mom.shape))
# PLUS_DI - Plus Directional Indicator
plusdi = talib.PLUS_DI(high, low, close, timeperiod=14)
plusdi = plusdi.fillna(plusdi.mean())
print('PLUS_DI - Plus Directional Indicator shape: {}'.format(
plusdi.shape))
# PLUS_DM - Plus Directional Movement
plusdm = talib.PLUS_DM(high, low, timeperiod=14)
plusdm = plusdm.fillna(plusdm.mean())
print('PLUS_DM - Plus Directional Movement shape: {}'.format(plusdi.shape))
# PPO - Percentage Price Oscillator
ppo = talib.PPO(close, fastperiod=12, slowperiod=26, matype=0)
ppo = ppo.fillna(ppo.mean())
print('PPO - Percentage Price Oscillator shape: {}'.format(ppo.shape))
# ROC - Rate of change:((price/prevPrice)-1)*100
roc = talib.ROC(close, timeperiod=10)
roc = roc.fillna(roc.mean())
print('ROC - Rate of change : ((price/prevPrice)-1)*100 shape: {}'.format(
roc.shape))
# RSI - Relative Strength Index
rsi = talib.RSI(close, timeperiod=14)
rsi = rsi.fillna(rsi.mean())
print('RSI - Relative Strength Index shape: {}'.format(rsi.shape))
# TRIX - 1-day Rate-Of-Change (ROC) of a Triple Smooth EMA
trix = talib.TRIX(close, timeperiod=30)
trix = trix.fillna(trix.mean())
print('TRIX - 1-day Rate-Of-Change (ROC) of a Triple Smooth EMA shape: {}'.
format(trix.shape))
# ULTOSC - Ultimate Oscillator
ultosc = talib.ULTOSC(high,
low,
close,
timeperiod1=7,
timeperiod2=14,
timeperiod3=28)
ultosc = ultosc.fillna(ultosc.mean())
print('ULTOSC - Ultimate Oscillator shape: {}'.format(ultosc.shape))
# WILLR - Williams'%R
willr = talib.WILLR(high, low, close, timeperiod=7)
willr = willr.fillna(willr.mean())
print("WILLR - Williams'%R shape: {}".format(willr.shape))
## Volume Indicator Functions
# AD - Chaikin A/D Line
ad = talib.AD(high, low, close, volume)
ad = ad.fillna(ad.mean())
print('AD - Chaikin A/D Line shape: {}'.format(ad.shape))
# ADOSC - Chaikin A/D Oscillator
adosc = talib.ADOSC(high, low, close, volume, fastperiod=3, slowperiod=10)
adosc = adosc.fillna(adosc.mean())
print('ADOSC - Chaikin A/D Oscillator shape: {}'.format(adosc.shape))
# OBV - On Balance Volume
obv = talib.OBV(close, volume)
obv = obv.fillna(obv.mean())
print('OBV - On Balance Volume shape: {}'.format(obv.shape))
## Volatility Indicator Functions
# ATR - Average True Range
atr = talib.ATR(high, low, close, timeperiod=7)
atr = atr.fillna(atr.mean())
print('ATR - Average True Range shape: {}'.format(atr.shape))
# NATR - Normalized Average True Range
natr = talib.NATR(high, low, close, timeperiod=7)
natr = natr.fillna(natr.mean())
print('NATR - Normalized Average True Range shape: {}'.format(natr.shape))
# TRANGE - True Range
trange = talib.TRANGE(high, low, close)
trange = trange.fillna(trange.mean())
print('TRANGE - True Range shape: {}'.format(natr.shape))
## Price Transform Functions
# AVGPRICE - Average Price
avg = talib.AVGPRICE(open, high, low, close)
avg = avg.fillna(avg.mean())
print('AVGPRICE - Average Price shape: {}'.format(natr.shape))
# MEDPRICE - Median Price
medprice = talib.MEDPRICE(high, low)
medprice = medprice.fillna(medprice.mean())
print('MEDPRICE - Median Price shape: {}'.format(medprice.shape))
# TYPPRICE - Typical Price
typ = talib.TYPPRICE(high, low, close)
typ = typ.fillna(typ.mean())
print('TYPPRICE - Typical Price shape: {}'.format(typ.shape))
# WCLPRICE - Weighted Close Price
wcl = talib.WCLPRICE(high, low, close)
wcl = wcl.fillna(wcl.mean())
print('WCLPRICE - Weighted Close Price shape: {}'.format(wcl.shape))
## Cycle Indicator Functions
# HT_DCPERIOD - Hilbert Transform - Dominant Cycle Period
dcperiod = talib.HT_DCPERIOD(close)
dcperiod = dcperiod.fillna(dcperiod.mean())
print('HT_DCPERIOD - Hilbert Transform - Dominant Cycle Period shape: {}'.
format(dcperiod.shape))
# HT_DCPHASE - Hilbert Transform - Dominant Cycle Phase
dcphase = talib.HT_DCPHASE(close)
dcphase = dcphase.fillna(dcphase.mean())
print('HT_DCPHASE - Hilbert Transform - Dominant Cycle Phase shape: {}'.
format(dcperiod.shape))
## Statistic Functions
# BETA - Beta
beta = talib.BETA(high, low, timeperiod=3)
beta = beta.fillna(beta.mean())
print('BETA - Beta shape: {}'.format(beta.shape))
# CORREL - Pearson's Correlation Coefficient(r)
correl = talib.CORREL(high, low, timeperiod=30)
correl = correl.fillna(correl.mean())
print("CORREL - Pearson's Correlation Coefficient(r) shape: {}".format(
beta.shape))
# LINEARREG - Linear Regression
linreg = talib.LINEARREG(close, timeperiod=7)
linreg = linreg.fillna(linreg.mean())
print("LINEARREG - Linear Regression shape: {}".format(linreg.shape))
# STDDEV - Standard Deviation
stddev = talib.STDDEV(close, timeperiod=5, nbdev=1)
stddev = stddev.fillna(stddev.mean())
print("STDDEV - Standard Deviation shape: {}".format(stddev.shape))
# TSF - Time Series Forecast
tsf = talib.TSF(close, timeperiod=7)
tsf = tsf.fillna(tsf.mean())
print("TSF - Time Series Forecast shape: {}".format(tsf.shape))
# VAR - Variance
var = talib.VAR(close, timeperiod=5, nbdev=1)
var = var.fillna(var.mean())
print("VAR - Variance shape: {}".format(var.shape))
## Feature DataFrame
X_full = pd.concat([
X, dema, ema, hilbert, kama, ma, midpoint, midprice, sar, sarext, sma,
t3, tema, trima, wma, adx, adxr, apo, aroon, bop, cci, cmo, mfi,
minusdi, minusdm, mom, plusdi, plusdm, ppo, roc, rsi, trix, ultosc,
willr, ad, adosc, obv, atr, natr, trange, avg, medprice, typ, wcl,
dcperiod, dcphase, beta, correl, linreg, stddev, tsf, var
],
axis=1).drop(['open', 'high', 'low', 'close'], axis=1)
X_full.columns = [
'volume', 'diff', 'dema', 'ema', 'hilbert', 'kama', 'ma', 'midpoint',
'midprice', 'sar', 'sarext', 'sma', 't3', 'tema', 'trima', 'wma',
'adx', 'adxr', 'apo', 'aroon', 'bop', 'cci', 'cmo', 'mfi', 'minusdi',
'minusdm', 'mom', 'plusdi', 'plusdm', 'ppo', 'roc', 'rsi', 'trix',
'ultosc', 'willr', 'ad', 'adosc', 'obv', 'atr', 'natr', 'trange',
'avg', 'medprice', 'typ', 'wcl', 'dcperiod', 'dcphase', 'beta',
'correl', 'linreg', 'stddev', 'tsf', 'var'
]
X_full.join(y_).head(10)
# full feature models
X_train_full, X_test_full, y_train_full, y_test_full = train_test_split(
X_full, y_, test_size=0.33, random_state=42)
print('X_train shape: {}'.format(X_train_full.shape))
print('X_test shape: {}'.format(X_test_full.shape))
print('y_train shape: {}'.format(y_train_full.shape))
print('y_test shape: {}'.format(y_test_full.shape))
pipe_knn_full = Pipeline([('scl', StandardScaler()),
('est', KNeighborsClassifier(n_neighbors=3))])
pipe_logistic_full = Pipeline([
('scl', StandardScaler()),
('est',
LogisticRegression(solver='lbfgs',
multi_class='multinomial',
random_state=39))
])
pipe_rf_full = Pipeline([('scl', StandardScaler()),
('est', RandomForestClassifier(random_state=39))])
pipe_gb_full = Pipeline([('scl', StandardScaler()),
('est',
GradientBoostingClassifier(random_state=39))])
pipe_names = ['KNN', 'Logistic', 'RandomForest', 'GradientBoosting']
pipe_lines_full = [
pipe_knn_full, pipe_logistic_full, pipe_rf_full, pipe_gb_full
]
for (i, pipe) in enumerate(pipe_lines_full):
pipe.fit(X_train_full, y_train_full.values.ravel())
print(pipe)
print('%s: %.3f' % (pipe_names[i] + ' Train Accuracy',
accuracy_score(y_train_full.values.ravel(),
pipe.predict(X_train_full))))
print('%s: %.3f' % (pipe_names[i] + ' Test Accuracy',
accuracy_score(y_test_full.values.ravel(),
pipe.predict(X_test_full))))
print('%s: %.3f' % (pipe_names[i] + ' Train F1 Score',
f1_score(y_train_full.values.ravel(),
pipe.predict(X_train_full),
average='micro')))
print('%s: %.3f' % (pipe_names[i] + ' Test F1 Score',
f1_score(y_test_full.values.ravel(),
pipe.predict(X_test_full),
average='micro')))
# Univariate Statistics
select = SelectPercentile(percentile=25)
select.fit(X_train_full, y_train_full.values.ravel())
X_train_selected = select.transform(X_train_full)
X_test_selected = select.transform(X_test_full)
# GradientBoost Classifier
print(
'--------------------------Without Univariate Statistics-------------------------------------'
)
pipe_gb = Pipeline([('scl', StandardScaler()),
('est', GradientBoostingClassifier(random_state=39))])
pipe_gb.fit(X_train_full, y_train_full.values.ravel())
print('Train Accuracy: {:.3f}'.format(
accuracy_score(y_train_full.values.ravel(),
pipe_gb.predict(X_train_full))))
print('Test Accuracy: {:.3f}'.format(
accuracy_score(y_test_full.values.ravel(),
pipe_gb.predict(X_test_full))))
print('Train F1 Score: {:.3f}'.format(
f1_score(y_train_full.values.ravel(),
pipe_gb.predict(X_train_full),
average='micro')))
print('Test F1 Score: {:.3f}'.format(
f1_score(y_test_full.values.ravel(),
pipe_gb.predict(X_test_full),
average='micro')))
# GradientBoost Cllassifier with Univariate Statistics
print(
'---------------------------With Univariate Statistics--------------------------------------'
)
pipe_gb_percentile = Pipeline([
('scl', StandardScaler()),
('est', GradientBoostingClassifier(random_state=39))
])
pipe_gb_percentile.fit(X_train_selected, y_train_full.values.ravel())
print('Train Accuracy: {:.3f}'.format(
accuracy_score(y_train_full.values.ravel(),
pipe_gb_percentile.predict(X_train_selected))))
print('Test Accuracy: {:.3f}'.format(
accuracy_score(y_test_full.values.ravel(),
pipe_gb_percentile.predict(X_test_selected))))
print('Train F1 Score: {:.3f}'.format(
f1_score(y_train_full.values.ravel(),
pipe_gb_percentile.predict(X_train_selected),
average='micro')))
print('Test F1 Score: {:.3f}'.format(
f1_score(y_test_full.values.ravel(),
pipe_gb_percentile.predict(X_test_selected),
average='micro')))
# Model-based Selection
select = SelectFromModel(RandomForestClassifier(n_estimators=100,
random_state=42),
threshold="1.25*mean")
select.fit(X_train_full, y_train_full.values.ravel())
X_train_model = select.transform(X_train_full)
X_test_model = select.transform(X_test_full)
# GradientBoost Classifier
print(
'--------------------------Without Model-based Selection--------------------------------------'
)
pipe_gb = Pipeline([('scl', StandardScaler()),
('est', GradientBoostingClassifier(random_state=39))])
pipe_gb.fit(X_train_full, y_train_full.values.ravel())
print('Train Accuracy: {:.3f}'.format(
accuracy_score(y_train_full.values.ravel(),
pipe_gb.predict(X_train_full))))
print('Test Accuracy: {:.3f}'.format(
accuracy_score(y_test_full.values.ravel(),
pipe_gb.predict(X_test_full))))
print('Train F1 Score: {:.3f}'.format(
f1_score(y_train_full.values.ravel(),
pipe_gb.predict(X_train_full),
average='micro')))
print('Test F1 Score: {:.3f}'.format(
f1_score(y_test_full.values.ravel(),
pipe_gb.predict(X_test_full),
average='micro')))
# GradientBoost Classifier with Model-based Selection
print(
'----------------------------With Model-based Selection--------------------------------------'
)
pipe_gb_model = Pipeline([('scl', StandardScaler()),
('est',
GradientBoostingClassifier(random_state=39))])
pipe_gb_model.fit(X_train_model, y_train_full.values.ravel())
print('Train Accuracy: {:.3f}'.format(
accuracy_score(y_train_full.values.ravel(),
pipe_gb_model.predict(X_train_model))))
print('Test Accuracy: {:.3f}'.format(
accuracy_score(y_test_full.values.ravel(),
pipe_gb_model.predict(X_test_model))))
print('Train F1 Score: {:.3f}'.format(
f1_score(y_train_full.values.ravel(),
pipe_gb_model.predict(X_train_model),
average='micro')))
print('Test F1 Score: {:.3f}'.format(
f1_score(y_test_full.values.ravel(),
pipe_gb_model.predict(X_test_model),
average='micro')))
# Recursive Feature Elimination
select = RFE(RandomForestClassifier(n_estimators=100, random_state=42),
n_features_to_select=15)
select.fit(X_train_full, y_train_full.values.ravel())
X_train_rfe = select.transform(X_train_full)
X_test_rfe = select.transform(X_test_full)
# GradientBoost Classifier
print(
'--------------------------Without Recursive Feature Elimination-------------------------------------'
)
pipe_gb = Pipeline([('scl', StandardScaler()),
('est', GradientBoostingClassifier(random_state=39))])
pipe_gb.fit(X_train_full, y_train_full.values.ravel())
print('Train Accuracy: {:.3f}'.format(
accuracy_score(y_train_full.values.ravel(),
pipe_gb.predict(X_train_full))))
print('Test Accuracy: {:.3f}'.format(
accuracy_score(y_test_full.values.ravel(),
pipe_gb.predict(X_test_full))))
print('Train F1 Score: {:.3f}'.format(
f1_score(y_train_full.values.ravel(),
pipe_gb.predict(X_train_full),
average='micro')))
print('Test F1 Score: {:.3f}'.format(
f1_score(y_test_full.values.ravel(),
pipe_gb.predict(X_test_full),
average='micro')))
# GradientBoost Classifier with Recursive Feature Elimination
print(
'----------------------------With Recursive Feature Elimination--------------------------------------'
)
pipe_gb_rfe = Pipeline([('scl', StandardScaler()),
('est',
GradientBoostingClassifier(random_state=39))])
pipe_gb_rfe.fit(X_train_rfe, y_train_full.values.ravel())
print('Train Accuracy: {:.3f}'.format(
accuracy_score(y_train_full.values.ravel(),
pipe_gb_rfe.predict(X_train_rfe))))
print('Test Accuracy: {:.3f}'.format(
accuracy_score(y_test_full.values.ravel(),
pipe_gb_rfe.predict(X_test_rfe))))
print('Train F1 Score: {:.3f}'.format(
f1_score(y_train_full.values.ravel(),
pipe_gb_rfe.predict(X_train_rfe),
average='micro')))
print('Test F1 Score: {:.3f}'.format(
f1_score(y_test_full.values.ravel(),
pipe_gb_rfe.predict(X_test_rfe),
average='micro')))
cv = cross_val_score(pipe_gb,
X_,
y_.values.ravel(),
cv=StratifiedKFold(n_splits=10,
shuffle=True,
random_state=39))
print('Cross validation with StratifiedKFold scores: {}'.format(cv))
print('Cross Validation with StatifiedKFold mean: {}'.format(cv.mean()))
# GridSearch
n_features = len(df.columns)
param_grid = {
'learning_rate': [0.01, 0.1, 1, 10],
'n_estimators': [1, 10, 100, 200, 300],
'max_depth': [1, 2, 3, 4, 5]
}
stratifiedcv = StratifiedKFold(n_splits=10, shuffle=True, random_state=39)
X_train, X_test, y_train, y_test = train_test_split(X_,
y_,
test_size=0.33,
random_state=42)
grid_search = GridSearchCV(GradientBoostingClassifier(),
param_grid,
cv=stratifiedcv)
grid_search.fit(X_train, y_train.values.ravel())
print('GridSearch Train Accuracy: {:.3f}'.format(
accuracy_score(y_train.values.ravel(), grid_search.predict(X_train))))
print('GridSearch Test Accuracy: {:.3f}'.format(
accuracy_score(y_test.values.ravel(), grid_search.predict(X_test))))
print('GridSearch Train F1 Score: {:.3f}'.format(
f1_score(y_train.values.ravel(),
grid_search.predict(X_train),
average='micro')))
print('GridSearch Test F1 Score: {:.3f}'.format(
f1_score(y_test.values.ravel(),
grid_search.predict(X_test),
average='micro')))
# GridSearch results
print("Best params:\n{}".format(grid_search.best_params_))
print("Best cross-validation score: {:.2f}".format(
grid_search.best_score_))
results = pd.DataFrame(grid_search.cv_results_)
corr_params = results.drop(results.columns[[
0, 1, 2, 3, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 19, 20
]],
axis=1)
corr_params.head()
# GridSearch in nested
cv_gb = cross_val_score(grid_search,
X_,
y_.values.ravel(),
cv=StratifiedKFold(n_splits=3,
shuffle=True,
random_state=39))
print('Grid Search with nested cross validation scores: {}'.format(cv_gb))
print('Grid Search with nested cross validation mean: {}'.format(
cv_gb.mean()))
def calculate(self, para):
self.t = self.inputdata[:, 0]
self.op = self.inputdata[:, 1]
self.high = self.inputdata[:, 2]
self.low = self.inputdata[:, 3]
self.close = self.inputdata[:, 4]
#adjusted close
self.close1 = self.inputdata[:, 5]
self.volume = self.inputdata[:, 6]
#Overlap study
#Overlap Studies
#Overlap Studies
if para is 'BBANDS': #Bollinger Bands
upperband, middleband, lowerband = ta.BBANDS(self.close,
timeperiod=self.tp,
nbdevup=2,
nbdevdn=2,
matype=0)
self.output = [upperband, middleband, lowerband]
elif para is 'DEMA': #Double Exponential Moving Average
self.output = ta.DEMA(self.close, timeperiod=self.tp)
elif para is 'EMA': #Exponential Moving Average
self.output = ta.EMA(self.close, timeperiod=self.tp)
elif para is 'HT_TRENDLINE': #Hilbert Transform - Instantaneous Trendline
self.output = ta.HT_TRENDLINE(self.close)
elif para is 'KAMA': #Kaufman Adaptive Moving Average
self.output = ta.KAMA(self.close, timeperiod=self.tp)
elif para is 'MA': #Moving average
self.output = ta.MA(self.close, timeperiod=self.tp, matype=0)
elif para is 'MAMA': #MESA Adaptive Moving Average
mama, fama = ta.MAMA(self.close, fastlimit=0, slowlimit=0)
elif para is 'MAVP': #Moving average with variable period
self.output = ta.MAVP(self.close,
periods=10,
minperiod=self.tp,
maxperiod=self.tp1,
matype=0)
elif para is 'MIDPOINT': #MidPoint over period
self.output = ta.MIDPOINT(self.close, timeperiod=self.tp)
elif para is 'MIDPRICE': #Midpoint Price over period
self.output = ta.MIDPRICE(self.high, self.low, timeperiod=self.tp)
elif para is 'SAR': #Parabolic SAR
self.output = ta.SAR(self.high,
self.low,
acceleration=0,
maximum=0)
elif para is 'SAREXT': #Parabolic SAR - Extended
self.output = ta.SAREXT(self.high,
self.low,
startvalue=0,
offsetonreverse=0,
accelerationinitlong=0,
accelerationlong=0,
accelerationmaxlong=0,
accelerationinitshort=0,
accelerationshort=0,
accelerationmaxshort=0)
elif para is 'SMA': #Simple Moving Average
self.output = ta.SMA(self.close, timeperiod=self.tp)
elif para is 'T3': #Triple Exponential Moving Average (T3)
self.output = ta.T3(self.close, timeperiod=self.tp, vfactor=0)
elif para is 'TEMA': #Triple Exponential Moving Average
self.output = ta.TEMA(self.close, timeperiod=self.tp)
elif para is 'TRIMA': #Triangular Moving Average
self.output = ta.TRIMA(self.close, timeperiod=self.tp)
elif para is 'WMA': #Weighted Moving Average
self.output = ta.WMA(self.close, timeperiod=self.tp)
#Momentum Indicators
elif para is 'ADX': #Average Directional Movement Index
self.output = ta.ADX(self.high,
self.low,
self.close,
timeperiod=self.tp)
elif para is 'ADXR': #Average Directional Movement Index Rating
self.output = ta.ADXR(self.high,
self.low,
self.close,
timeperiod=self.tp)
elif para is 'APO': #Absolute Price Oscillator
self.output = ta.APO(self.close,
fastperiod=12,
slowperiod=26,
matype=0)
elif para is 'AROON': #Aroon
aroondown, aroonup = ta.AROON(self.high,
self.low,
timeperiod=self.tp)
self.output = [aroondown, aroonup]
elif para is 'AROONOSC': #Aroon Oscillator
self.output = ta.AROONOSC(self.high, self.low, timeperiod=self.tp)
elif para is 'BOP': #Balance Of Power
self.output = ta.BOP(self.op, self.high, self.low, self.close)
elif para is 'CCI': #Commodity Channel Index
self.output = ta.CCI(self.high,
self.low,
self.close,
timeperiod=self.tp)
elif para is 'CMO': #Chande Momentum Oscillator
self.output = ta.CMO(self.close, timeperiod=self.tp)
elif para is 'DX': #Directional Movement Index
self.output = ta.DX(self.high,
self.low,
self.close,
timeperiod=self.tp)
elif para is 'MACD': #Moving Average Convergence/Divergence
macd, macdsignal, macdhist = ta.MACD(self.close,
fastperiod=12,
slowperiod=26,
signalperiod=9)
self.output = [macd, macdsignal, macdhist]
elif para is 'MACDEXT': #MACD with controllable MA type
macd, macdsignal, macdhist = ta.MACDEXT(self.close,
fastperiod=12,
fastmatype=0,
slowperiod=26,
slowmatype=0,
signalperiod=9,
signalmatype=0)
self.output = [macd, macdsignal, macdhist]
elif para is 'MACDFIX': #Moving Average Convergence/Divergence Fix 12/26
macd, macdsignal, macdhist = ta.MACDFIX(self.close, signalperiod=9)
self.output = [macd, macdsignal, macdhist]
elif para is 'MFI': #Money Flow Index
self.output = ta.MFI(self.high,
self.low,
self.close,
self.volume,
timeperiod=self.tp)
elif para is 'MINUS_DI': #Minus Directional Indicator
self.output = ta.MINUS_DI(self.high,
self.low,
self.close,
timeperiod=self.tp)
elif para is 'MINUS_DM': #Minus Directional Movement
self.output = ta.MINUS_DM(self.high, self.low, timeperiod=self.tp)
elif para is 'MOM': #Momentum
self.output = ta.MOM(self.close, timeperiod=10)
elif para is 'PLUS_DI': #Plus Directional Indicator
self.output = ta.PLUS_DI(self.high,
self.low,
self.close,
timeperiod=self.tp)
elif para is 'PLUS_DM': #Plus Directional Movement
self.output = ta.PLUS_DM(self.high, self.low, timeperiod=self.tp)
elif para is 'PPO': #Percentage Price Oscillator
self.output = ta.PPO(self.close,
fastperiod=12,
slowperiod=26,
matype=0)
elif para is 'ROC': #Rate of change : ((price/prevPrice)-1)*100
self.output = ta.ROC(self.close, timeperiod=10)
elif para is 'ROCP': #Rate of change Percentage: (price-prevPrice)/prevPrice
self.output = ta.ROCP(self.close, timeperiod=10)
elif para is 'ROCR': #Rate of change ratio: (price/prevPrice)
self.output = ta.ROCR(self.close, timeperiod=10)
elif para is 'ROCR100': #Rate of change ratio 100 scale: (price/prevPrice)*100
self.output = ta.ROCR100(self.close, timeperiod=10)
elif para is 'RSI': #Relative Strength Index
self.output = ta.RSI(self.close, timeperiod=self.tp)
elif para is 'STOCH': #Stochastic
slowk, slowd = ta.STOCH(self.high,
self.low,
self.close,
fastk_period=5,
slowk_period=3,
slowk_matype=0,
slowd_period=3,
slowd_matype=0)
self.output = [slowk, slowd]
elif para is 'STOCHF': #Stochastic Fast
fastk, fastd = ta.STOCHF(self.high,
self.low,
self.close,
fastk_period=5,
fastd_period=3,
fastd_matype=0)
self.output = [fastk, fastd]
elif para is 'STOCHRSI': #Stochastic Relative Strength Index
fastk, fastd = ta.STOCHRSI(self.close,
timeperiod=self.tp,
fastk_period=5,
fastd_period=3,
fastd_matype=0)
self.output = [fastk, fastd]
elif para is 'TRIX': #1-day Rate-Of-Change (ROC) of a Triple Smooth EMA
self.output = ta.TRIX(self.close, timeperiod=self.tp)
elif para is 'ULTOSC': #Ultimate Oscillator
self.output = ta.ULTOSC(self.high,
self.low,
self.close,
timeperiod1=self.tp,
timeperiod2=self.tp1,
timeperiod3=self.tp2)
elif para is 'WILLR': #Williams' %R
self.output = ta.WILLR(self.high,
self.low,
self.close,
timeperiod=self.tp)
# Volume Indicators : #
elif para is 'AD': #Chaikin A/D Line
self.output = ta.AD(self.high, self.low, self.close, self.volume)
elif para is 'ADOSC': #Chaikin A/D Oscillator
self.output = ta.ADOSC(self.high,
self.low,
self.close,
self.volume,
fastperiod=3,
slowperiod=10)
elif para is 'OBV': #On Balance Volume
self.output = ta.OBV(self.close, self.volume)
# Volatility Indicators: #
elif para is 'ATR': #Average True Range
self.output = ta.ATR(self.high,
self.low,
self.close,
timeperiod=self.tp)
elif para is 'NATR': #Normalized Average True Range
self.output = ta.NATR(self.high,
self.low,
self.close,
timeperiod=self.tp)
elif para is 'TRANGE': #True Range
self.output = ta.TRANGE(self.high, self.low, self.close)
#Price Transform : #
elif para is 'AVGPRICE': #Average Price
self.output = ta.AVGPRICE(self.op, self.high, self.low, self.close)
elif para is 'MEDPRICE': #Median Price
self.output = ta.MEDPRICE(self.high, self.low)
elif para is 'TYPPRICE': #Typical Price
self.output = ta.TYPPRICE(self.high, self.low, self.close)
elif para is 'WCLPRICE': #Weighted Close Price
self.output = ta.WCLPRICE(self.high, self.low, self.close)
#Cycle Indicators : #
elif para is 'HT_DCPERIOD': #Hilbert Transform - Dominant Cycle Period
self.output = ta.HT_DCPERIOD(self.close)
elif para is 'HT_DCPHASE': #Hilbert Transform - Dominant Cycle Phase
self.output = ta.HT_DCPHASE(self.close)
elif para is 'HT_PHASOR': #Hilbert Transform - Phasor Components
inphase, quadrature = ta.HT_PHASOR(self.close)
self.output = [inphase, quadrature]
elif para is 'HT_SINE': #Hilbert Transform - SineWave #2
sine, leadsine = ta.HT_SINE(self.close)
self.output = [sine, leadsine]
elif para is 'HT_TRENDMODE': #Hilbert Transform - Trend vs Cycle Mode
self.integer = ta.HT_TRENDMODE(self.close)
#Pattern Recognition : #
elif para is 'CDL2CROWS': #Two Crows
self.integer = ta.CDL2CROWS(self.op, self.high, self.low,
self.close)
elif para is 'CDL3BLACKCROWS': #Three Black Crows
self.integer = ta.CDL3BLACKCROWS(self.op, self.high, self.low,
self.close)
elif para is 'CDL3INSIDE': #Three Inside Up/Down
self.integer = ta.CDL3INSIDE(self.op, self.high, self.low,
self.close)
elif para is 'CDL3LINESTRIKE': #Three-Line Strike
self.integer = ta.CDL3LINESTRIKE(self.op, self.high, self.low,
self.close)
elif para is 'CDL3OUTSIDE': #Three Outside Up/Down
self.integer = ta.CDL3OUTSIDE(self.op, self.high, self.low,
self.close)
elif para is 'CDL3STARSINSOUTH': #Three Stars In The South
self.integer = ta.CDL3STARSINSOUTH(self.op, self.high, self.low,
self.close)
elif para is 'CDL3WHITESOLDIERS': #Three Advancing White Soldiers
self.integer = ta.CDL3WHITESOLDIERS(self.op, self.high, self.low,
self.close)
elif para is 'CDLABANDONEDBABY': #Abandoned Baby
self.integer = ta.CDLABANDONEDBABY(self.op,
self.high,
self.low,
self.close,
penetration=0)
elif para is 'CDLBELTHOLD': #Belt-hold
self.integer = ta.CDLBELTHOLD(self.op, self.high, self.low,
self.close)
elif para is 'CDLBREAKAWAY': #Breakaway
self.integer = ta.CDLBREAKAWAY(self.op, self.high, self.low,
self.close)
elif para is 'CDLCLOSINGMARUBOZU': #Closing Marubozu
self.integer = ta.CDLCLOSINGMARUBOZU(self.op, self.high, self.low,
self.close)
elif para is 'CDLCONCEALBABYSWALL': #Concealing Baby Swallow
self.integer = ta.CDLCONCEALBABYSWALL(self.op, self.high, self.low,
self.close)
elif para is 'CDLCOUNTERATTACK': #Counterattack
self.integer = ta.CDLCOUNTERATTACK(self.op, self.high, self.low,
self.close)
elif para is 'CDLDARKCLOUDCOVER': #Dark Cloud Cover
self.integer = ta.CDLDARKCLOUDCOVER(self.op,
self.high,
self.low,
self.close,
penetration=0)
elif para is 'CDLDOJI': #Doji
self.integer = ta.CDLDOJI(self.op, self.high, self.low, self.close)
elif para is 'CDLDOJISTAR': #Doji Star
self.integer = ta.CDLDOJISTAR(self.op, self.high, self.low,
self.close)
elif para is 'CDLDRAGONFLYDOJI': #Dragonfly Doji
self.integer = ta.CDLDRAGONFLYDOJI(self.op, self.high, self.low,
self.close)
elif para is 'CDLENGULFING': #Engulfing Pattern
self.integer = ta.CDLENGULFING(self.op, self.high, self.low,
self.close)
elif para is 'CDLEVENINGDOJISTAR': #Evening Doji Star
self.integer = ta.CDLEVENINGDOJISTAR(self.op,
self.high,
self.low,
self.close,
penetration=0)
elif para is 'CDLEVENINGSTAR': #Evening Star
self.integer = ta.CDLEVENINGSTAR(self.op,
self.high,
self.low,
self.close,
penetration=0)
elif para is 'CDLGAPSIDESIDEWHITE': #Up/Down-gap side-by-side white lines
self.integer = ta.CDLGAPSIDESIDEWHITE(self.op, self.high, self.low,
self.close)
elif para is 'CDLGRAVESTONEDOJI': #Gravestone Doji
self.integer = ta.CDLGRAVESTONEDOJI(self.op, self.high, self.low,
self.close)
elif para is 'CDLHAMMER': #Hammer
self.integer = ta.CDLHAMMER(self.op, self.high, self.low,
self.close)
elif para is 'CDLHANGINGMAN': #Hanging Man
self.integer = ta.CDLHANGINGMAN(self.op, self.high, self.low,
self.close)
elif para is 'CDLHARAMI': #Harami Pattern
self.integer = ta.CDLHARAMI(self.op, self.high, self.low,
self.close)
elif para is 'CDLHARAMICROSS': #Harami Cross Pattern
self.integer = ta.CDLHARAMICROSS(self.op, self.high, self.low,
self.close)
elif para is 'CDLHIGHWAVE': #High-Wave Candle
self.integer = ta.CDLHIGHWAVE(self.op, self.high, self.low,
self.close)
elif para is 'CDLHIKKAKE': #Hikkake Pattern
self.integer = ta.CDLHIKKAKE(self.op, self.high, self.low,
self.close)
elif para is 'CDLHIKKAKEMOD': #Modified Hikkake Pattern
self.integer = ta.CDLHIKKAKEMOD(self.op, self.high, self.low,
self.close)
elif para is 'CDLHOMINGPIGEON': #Homing Pigeon
self.integer = ta.CDLHOMINGPIGEON(self.op, self.high, self.low,
self.close)
elif para is 'CDLIDENTICAL3CROWS': #Identical Three Crows
self.integer = ta.CDLIDENTICAL3CROWS(self.op, self.high, self.low,
self.close)
elif para is 'CDLINNECK': #In-Neck Pattern
self.integer = ta.CDLINNECK(self.op, self.high, self.low,
self.close)
elif para is 'CDLINVERTEDHAMMER': #Inverted Hammer
self.integer = ta.CDLINVERTEDHAMMER(self.op, self.high, self.low,
self.close)
elif para is 'CDLKICKING': #Kicking
self.integer = ta.CDLKICKING(self.op, self.high, self.low,
self.close)
elif para is 'CDLKICKINGBYLENGTH': #Kicking - bull/bear determined by the longer marubozu
self.integer = ta.CDLKICKINGBYLENGTH(self.op, self.high, self.low,
self.close)
elif para is 'CDLLADDERBOTTOM': #Ladder Bottom
self.integer = ta.CDLLADDERBOTTOM(self.op, self.high, self.low,
self.close)
elif para is 'CDLLONGLEGGEDDOJI': #Long Legged Doji
self.integer = ta.CDLLONGLEGGEDDOJI(self.op, self.high, self.low,
self.close)
elif para is 'CDLLONGLINE': #Long Line Candle
self.integer = ta.CDLLONGLINE(self.op, self.high, self.low,
self.close)
elif para is 'CDLMARUBOZU': #Marubozu
self.integer = ta.CDLMARUBOZU(self.op, self.high, self.low,
self.close)
elif para is 'CDLMATCHINGLOW': #Matching Low
self.integer = ta.CDLMATCHINGLOW(self.op, self.high, self.low,
self.close)
elif para is 'CDLMATHOLD': #Mat Hold
self.integer = ta.CDLMATHOLD(self.op,
self.high,
self.low,
self.close,
penetration=0)
elif para is 'CDLMORNINGDOJISTAR': #Morning Doji Star
self.integer = ta.CDLMORNINGDOJISTAR(self.op,
self.high,
self.low,
self.close,
penetration=0)
elif para is 'CDLMORNINGSTAR': #Morning Star
self.integer = ta.CDLMORNINGSTAR(self.op,
self.high,
self.low,
self.close,
penetration=0)
elif para is 'CDLONNECK': #On-Neck Pattern
self.integer = ta.CDLONNECK(self.op, self.high, self.low,
self.close)
elif para is 'CDLPIERCING': #Piercing Pattern
self.integer = ta.CDLPIERCING(self.op, self.high, self.low,
self.close)
elif para is 'CDLRICKSHAWMAN': #Rickshaw Man
self.integer = ta.CDLRICKSHAWMAN(self.op, self.high, self.low,
self.close)
elif para is 'CDLRISEFALL3METHODS': #Rising/Falling Three Methods
self.integer = ta.CDLRISEFALL3METHODS(self.op, self.high, self.low,
self.close)
elif para is 'CDLSEPARATINGLINES': #Separating Lines
self.integer = ta.CDLSEPARATINGLINES(self.op, self.high, self.low,
self.close)
elif para is 'CDLSHOOTINGSTAR': #Shooting Star
self.integer = ta.CDLSHOOTINGSTAR(self.op, self.high, self.low,
self.close)
elif para is 'CDLSHORTLINE': #Short Line Candle
self.integer = ta.CDLSHORTLINE(self.op, self.high, self.low,
self.close)
elif para is 'CDLSPINNINGTOP': #Spinning Top
self.integer = ta.CDLSPINNINGTOP(self.op, self.high, self.low,
self.close)
elif para is 'CDLSTALLEDPATTERN': #Stalled Pattern
self.integer = ta.CDLSTALLEDPATTERN(self.op, self.high, self.low,
self.close)
elif para is 'CDLSTICKSANDWICH': #Stick Sandwich
self.integer = ta.CDLSTICKSANDWICH(self.op, self.high, self.low,
self.close)
elif para is 'CDLTAKURI': #Takuri (Dragonfly Doji with very long lower shadow)
self.integer = ta.CDLTAKURI(self.op, self.high, self.low,
self.close)
elif para is 'CDLTASUKIGAP': #Tasuki Gap
self.integer = ta.CDLTASUKIGAP(self.op, self.high, self.low,
self.close)
elif para is 'CDLTHRUSTING': #Thrusting Pattern
self.integer = ta.CDLTHRUSTING(self.op, self.high, self.low,
self.close)
elif para is 'CDLTRISTAR': #Tristar Pattern
self.integer = ta.CDLTRISTAR(self.op, self.high, self.low,
self.close)
elif para is 'CDLUNIQUE3RIVER': #Unique 3 River
self.integer = ta.CDLUNIQUE3RIVER(self.op, self.high, self.low,
self.close)
elif para is 'CDLUPSIDEGAP2CROWS': #Upside Gap Two Crows
self.integer = ta.CDLUPSIDEGAP2CROWS(self.op, self.high, self.low,
self.close)
elif para is 'CDLXSIDEGAP3METHODS': #Upside/Downside Gap Three Methods
self.integer = ta.CDLXSIDEGAP3METHODS(self.op, self.high, self.low,
self.close)
#Statistic Functions : #
elif para is 'BETA': #Beta
self.output = ta.BETA(self.high, self.low, timeperiod=5)
elif para is 'CORREL': #Pearson's Correlation Coefficient (r)
self.output = ta.CORREL(self.high, self.low, timeperiod=self.tp)
elif para is 'LINEARREG': #Linear Regression
self.output = ta.LINEARREG(self.close, timeperiod=self.tp)
elif para is 'LINEARREG_ANGLE': #Linear Regression Angle
self.output = ta.LINEARREG_ANGLE(self.close, timeperiod=self.tp)
elif para is 'LINEARREG_INTERCEPT': #Linear Regression Intercept
self.output = ta.LINEARREG_INTERCEPT(self.close,
timeperiod=self.tp)
elif para is 'LINEARREG_SLOPE': #Linear Regression Slope
self.output = ta.LINEARREG_SLOPE(self.close, timeperiod=self.tp)
elif para is 'STDDEV': #Standard Deviation
self.output = ta.STDDEV(self.close, timeperiod=5, nbdev=1)
elif para is 'TSF': #Time Series Forecast
self.output = ta.TSF(self.close, timeperiod=self.tp)
elif para is 'VAR': #Variance
self.output = ta.VAR(self.close, timeperiod=5, nbdev=1)
else:
print('You issued command:' + para)
def TALIB_ULTOSC(close, timeperiod1=7, timeperiod2=14, timeperiod3=28):
'''00393,4,1'''
return talib.ULTOSC(close, timeperiod1, timeperiod2, timeperiod3)
def process_data(self):
self.main_df = self.df.copy()
self.main_df["return_log"] = self.main_df["mid"].pct_change(1)
self.main_df["Close"] = talib.DEMA(
np.array(self.main_df["Close"]), timeperiod=3)
self.main_df["High"] = talib.DEMA(
np.array(self.main_df["High"]), timeperiod=3)
self.main_df["Low"] = talib.DEMA(
np.array(self.main_df["Low"]), timeperiod=3)
self.main_df["Open"] = talib.DEMA(
np.array(self.main_df["Open"]), timeperiod=3)
self.main_df["dema"] = talib.DEMA(
np.array(self.main_df["Close"]), timeperiod=30)
self.main_df["kama"] = talib.KAMA(np.array(
self.main_df["Close"]), timeperiod=30) # Kaufman Adaptive Moving Average
self.main_df["trima"] = talib.TRIMA(
np.array(self.main_df["Close"]), timeperiod=30) # Triple exponential
self.main_df["WMA"] = talib.WMA(
np.array(self.main_df["Close"]), timeperiod=30) # Weighted moving average
self.main_df["adx"] = talib.ADX(np.array(self.main_df["High"]), np.array(
self.main_df["Low"]), np.array(self.main_df["Close"]), timeperiod=14)
self.main_df["adxr"] = talib.ADXR(np.array(self.main_df["High"]), np.array(
self.main_df["Low"]), np.array(self.main_df["Close"]), timeperiod=14)
self.main_df["apo"] = talib.APO(
np.array(self.main_df["Close"]), fastperiod=12, slowperiod=26)
self.main_df["aroondown"], self.main_df["aroonup"] = talib.AROON(
np.array(self.main_df["High"]), np.array(self.main_df["Low"]), timeperiod=14)
self.main_df["aroonosc"] = talib.AROONOSC(
np.array(self.main_df["High"]), np.array(self.main_df["Low"]), timeperiod=14)
self.main_df["bop"] = talib.BOP(np.array(self.main_df["Open"]), np.array(
self.main_df["High"]), np.array(self.main_df["Low"]), np.array(self.main_df["Close"]))
self.main_df["cmo"] = talib.CMO(
np.array(self.main_df["Close"]), timeperiod=14)
self.main_df["dx"] = talib.DX(np.array(self.main_df["High"]), np.array(
self.main_df["Low"]), np.array(self.main_df["Close"]), timeperiod=14)
self.main_df["mfi"] = talib.MFI(np.array(self.main_df["High"]), np.array(self.main_df["Low"]), np.array(
self.main_df["Close"]), np.array(self.main_df["Volume"]), timeperiod=14)
self.main_df["minus_di"] = talib.MINUS_DI(np.array(self.main_df["High"]), np.array(
self.main_df["Low"]), np.array(self.main_df["Close"]), timeperiod=14)
self.main_df["minus_dm"] = talib.MINUS_DM(
np.array(self.main_df["High"]), np.array(self.main_df["Low"]), timeperiod=14)
self.main_df["plus_di"] = talib.PLUS_DI(np.array(self.main_df["High"]), np.array(
self.main_df["Low"]), np.array(self.main_df["Close"]), timeperiod=14)
self.main_df["plus_dm"] = talib.PLUS_DM(
np.array(self.main_df["High"]), np.array(self.main_df["Low"]), timeperiod=14)
self.main_df["ppo"] = talib.PPO(
np.array(self.main_df["Close"]), fastperiod=10, slowperiod=20)
self.main_df["rsi_14"] = talib.RSI(
np.array(self.main_df["Close"]), timeperiod=14)
self.main_df["slowk"], self.main_df["slowd"] = talib.STOCH(np.array(self.main_df["High"]), np.array(self.main_df["Low"]), np.array(
self.main_df["Close"]), fastk_period=5, slowk_period=3, slowk_matype=0, slowd_period=3, slowd_matype=0)
self.main_df["macd"], self.main_df["macdsignal"], self.main_df["macdhist"] = talib.MACD(
np.array(self.main_df["Close"]), fastperiod=12, slowperiod=26, signalperiod=9)
self.main_df["cci"] = talib.CCI(np.array(self.main_df["High"]), np.array(
self.main_df["Low"]), np.array(self.main_df["Close"]), timeperiod=14)
self.main_df["mom20"] = talib.MOM(
np.array(self.main_df["Close"]), timeperiod=20)
self.main_df["mom10"] = talib.MOM(
np.array(self.main_df["Close"]), timeperiod=10)
self.main_df["ma20"] = talib.SMA(
np.array(self.main_df["Close"]), timeperiod=20)
self.main_df["ma10"] = talib.SMA(
np.array(self.main_df["Close"]), timeperiod=10)
self.main_df["roc"] = talib.ROC(
np.array(self.main_df["Close"]), timeperiod=10)
self.main_df["ult"] = talib.ULTOSC(np.array(self.main_df["High"]), np.array(
self.main_df["Low"]), np.array(self.main_df["Close"]), timeperiod1=7, timeperiod2=14, timeperiod3=28)
self.main_df["will"] = talib.WILLR(np.array(self.main_df["High"]), np.array(
self.main_df["Low"]), np.array(self.main_df["Close"]), timeperiod=14)
self.main_df["return_1df"] = self.main_df["return_log"].shift(
-1)
self.main_df = self.main_df.dropna().iloc[140:, :]
return self.main_df
def calc_features(df):
open = df['op']
high = df['hi']
low = df['lo']
close = df['cl']
volume = df['volume']
orig_columns = df.columns
hilo = (df['hi'] + df['lo']) / 2
df['BBANDS_upperband'], df['BBANDS_middleband'], df[
'BBANDS_lowerband'] = talib.BBANDS(close,
timeperiod=5,
nbdevup=2,
nbdevdn=2,
matype=0)
df['BBANDS_upperband'] -= hilo
df['BBANDS_middleband'] -= hilo
df['BBANDS_lowerband'] -= hilo
df['DEMA'] = talib.DEMA(close, timeperiod=30) - hilo
df['EMA'] = talib.EMA(close, timeperiod=30) - hilo
df['HT_TRENDLINE'] = talib.HT_TRENDLINE(close) - hilo
df['KAMA'] = talib.KAMA(close, timeperiod=30) - hilo
df['MA'] = talib.MA(close, timeperiod=30, matype=0) - hilo
df['MIDPOINT'] = talib.MIDPOINT(close, timeperiod=14) - hilo
df['SMA'] = talib.SMA(close, timeperiod=30) - hilo
df['T3'] = talib.T3(close, timeperiod=5, vfactor=0) - hilo
df['TEMA'] = talib.TEMA(close, timeperiod=30) - hilo
df['TRIMA'] = talib.TRIMA(close, timeperiod=30) - hilo
df['WMA'] = talib.WMA(close, timeperiod=30) - hilo
df['ADX'] = talib.ADX(high, low, close, timeperiod=14)
df['ADXR'] = talib.ADXR(high, low, close, timeperiod=14)
df['APO'] = talib.APO(close, fastperiod=12, slowperiod=26, matype=0)
df['AROON_aroondown'], df['AROON_aroonup'] = talib.AROON(high,
low,
timeperiod=14)
df['AROONOSC'] = talib.AROONOSC(high, low, timeperiod=14)
df['BOP'] = talib.BOP(open, high, low, close)
df['CCI'] = talib.CCI(high, low, close, timeperiod=14)
df['DX'] = talib.DX(high, low, close, timeperiod=14)
df['MACD_macd'], df['MACD_macdsignal'], df['MACD_macdhist'] = talib.MACD(
close, fastperiod=12, slowperiod=26, signalperiod=9)
# skip MACDEXT MACDFIX たぶん同じなので
df['MFI'] = talib.MFI(high, low, close, volume, timeperiod=14)
df['MINUS_DI'] = talib.MINUS_DI(high, low, close, timeperiod=14)
df['MINUS_DM'] = talib.MINUS_DM(high, low, timeperiod=14)
df['MOM'] = talib.MOM(close, timeperiod=10)
df['PLUS_DI'] = talib.PLUS_DI(high, low, close, timeperiod=14)
df['PLUS_DM'] = talib.PLUS_DM(high, low, timeperiod=14)
df['RSI'] = talib.RSI(close, timeperiod=14)
df['STOCH_slowk'], df['STOCH_slowd'] = talib.STOCH(high,
low,
close,
fastk_period=5,
slowk_period=3,
slowk_matype=0,
slowd_period=3,
slowd_matype=0)
df['STOCHF_fastk'], df['STOCHF_fastd'] = talib.STOCHF(high,
low,
close,
fastk_period=5,
fastd_period=3,
fastd_matype=0)
df['STOCHRSI_fastk'], df['STOCHRSI_fastd'] = talib.STOCHRSI(close,
timeperiod=14,
fastk_period=5,
fastd_period=3,
fastd_matype=0)
df['TRIX'] = talib.TRIX(close, timeperiod=30)
df['ULTOSC'] = talib.ULTOSC(high,
low,
close,
timeperiod1=7,
timeperiod2=14,
timeperiod3=28)
df['WILLR'] = talib.WILLR(high, low, close, timeperiod=14)
df['AD'] = talib.AD(high, low, close, volume)
df['ADOSC'] = talib.ADOSC(high,
low,
close,
volume,
fastperiod=3,
slowperiod=10)
df['OBV'] = talib.OBV(close, volume)
df['ATR'] = talib.ATR(high, low, close, timeperiod=14)
df['NATR'] = talib.NATR(high, low, close, timeperiod=14)
df['TRANGE'] = talib.TRANGE(high, low, close)
df['HT_DCPERIOD'] = talib.HT_DCPERIOD(close)
df['HT_DCPHASE'] = talib.HT_DCPHASE(close)
df['HT_PHASOR_inphase'], df['HT_PHASOR_quadrature'] = talib.HT_PHASOR(
close)
df['HT_SINE_sine'], df['HT_SINE_leadsine'] = talib.HT_SINE(close)
df['HT_TRENDMODE'] = talib.HT_TRENDMODE(close)
df['BETA'] = talib.BETA(high, low, timeperiod=5)
df['CORREL'] = talib.CORREL(high, low, timeperiod=30)
df['LINEARREG'] = talib.LINEARREG(close, timeperiod=14) - close
df['LINEARREG_ANGLE'] = talib.LINEARREG_ANGLE(close, timeperiod=14)
df['LINEARREG_INTERCEPT'] = talib.LINEARREG_INTERCEPT(
close, timeperiod=14) - close
df['LINEARREG_SLOPE'] = talib.LINEARREG_SLOPE(close, timeperiod=14)
df['STDDEV'] = talib.STDDEV(close, timeperiod=5, nbdev=1)
return df
def get_datasets(asset, currency, granularity, datapoints):
"""Fetch the API and precess the desired pair
Arguments:
asset {str} -- First pair
currency {str} -- Second pair
granularity {str ['day', 'hour']} -- Granularity
datapoints {int [100 - 2000]} -- [description]
Returns:
pandas.Dataframe -- The OHLCV and indicators dataframe
"""
df_train_path = 'datasets/bot_train_{}_{}_{}.csv'.format(
asset + currency, datapoints, granularity)
df_rollout_path = 'datasets/bot_rollout_{}_{}_{}.csv'.format(
asset + currency, datapoints, granularity)
emojis = [
':moneybag:', ':yen:', ':dollar:', ':pound:', ':euro:',
':credit_card:', ':money_with_wings:', ':gem:'
]
if not os.path.exists(df_rollout_path):
headers = {
'User-Agent':
'Mozilla/5.0',
'authorization':
'Apikey 3d7d3e9e6006669ac00584978342451c95c3c78421268ff7aeef69995f9a09ce'
}
# OHLC
# url = 'https://min-api.cryptocompare.com/data/histo{}?fsym={}&tsym={}&e=Binance&limit={}'.format(granularity, asset, currency, datapoints)
url = 'https://min-api.cryptocompare.com/data/histo{}?fsym={}&tsym={}&limit={}'.format(
granularity, asset, currency, datapoints)
# print(emoji.emojize(':dizzy: :large_blue_diamond: :gem: :bar_chart: :crystal_ball: :chart_with_downwards_trend: :chart_with_upwards_trend: :large_orange_diamond: loading...', use_aliases=True))
print(
colored(
emoji.emojize('> ' + random.choice(emojis) + ' downloading ' +
asset + '/' + currency,
use_aliases=True), 'green'))
# print(colored('> downloading ' + asset + '/' + currency, 'green'))
response = requests.get(url, headers=headers)
json_response = response.json()
status = json_response['Response']
if status == "Error":
print(colored('=== {} ==='.format(json_response['Message']),
'red'))
raise AssertionError()
result = json_response['Data']
df = pd.DataFrame(result)
print(df.tail())
df['Date'] = pd.to_datetime(df['time'], utc=True, unit='s')
df.drop('time', axis=1, inplace=True)
# indicators
# https://github.com/mrjbq7/ta-lib/blob/master/docs/func.md
open_price, high, low, close = np.array(df['open']), np.array(
df['high']), np.array(df['low']), np.array(df['close'])
volume = np.array(df['volumefrom'])
# cycle indicators
df.loc[:, 'HT_DCPERIOD'] = talib.HT_DCPERIOD(close)
df.loc[:, 'HT_DCPHASE'] = talib.HT_DCPHASE(close)
df.loc[:,
'HT_PHASOR_inphase'], df.loc[:,
'HT_PHASOR_quadrature'] = talib.HT_PHASOR(
close)
df.loc[:, 'HT_SINE_sine'], df.loc[:,
'HT_SINE_leadsine'] = talib.HT_SINE(
close)
df.loc[:, 'HT_TRENDMODE'] = talib.HT_TRENDMODE(close)
# momemtum indicators
df.loc[:, 'ADX'] = talib.ADX(high, low, close, timeperiod=12)
df.loc[:, 'ADXR'] = talib.ADXR(high, low, close, timeperiod=13)
df.loc[:, 'APO'] = talib.APO(close,
fastperiod=5,
slowperiod=10,
matype=0)
df.loc[:,
'AROON_down'], df.loc[:,
'AROON_up'] = talib.AROON(high,
low,
timeperiod=15)
df.loc[:, 'AROONOSC'] = talib.AROONOSC(high, low, timeperiod=13)
df.loc[:, 'BOP'] = talib.BOP(open_price, high, low, close)
df.loc[:, 'CCI'] = talib.CCI(high, low, close, timeperiod=13)
df.loc[:, 'CMO'] = talib.CMO(close, timeperiod=14)
df.loc[:, 'DX'] = talib.DX(high, low, close, timeperiod=10)
df['MACD'], df['MACD_signal'], df['MACD_hist'] = talib.MACD(
close, fastperiod=5, slowperiod=10, signalperiod=20)
df.loc[:, 'MFI'] = talib.MFI(high, low, close, volume, timeperiod=12)
df.loc[:, 'MINUS_DI'] = talib.MINUS_DI(high, low, close, timeperiod=10)
df.loc[:, 'MINUS_DM'] = talib.MINUS_DM(high, low, timeperiod=14)
df.loc[:, 'MOM'] = talib.MOM(close, timeperiod=20)
df.loc[:, 'PPO'] = talib.PPO(close,
fastperiod=17,
slowperiod=35,
matype=2)
df.loc[:, 'ROC'] = talib.ROC(close, timeperiod=12)
df.loc[:, 'RSI'] = talib.RSI(close, timeperiod=25)
df.loc[:, 'STOCH_k'], df.loc[:,
'STOCH_d'] = talib.STOCH(high,
low,
close,
fastk_period=35,
slowk_period=12,
slowk_matype=0,
slowd_period=7,
slowd_matype=0)
df.loc[:,
'STOCHF_k'], df.loc[:,
'STOCHF_d'] = talib.STOCHF(high,
low,
close,
fastk_period=28,
fastd_period=14,
fastd_matype=0)
df.loc[:, 'STOCHRSI_K'], df.loc[:, 'STOCHRSI_D'] = talib.STOCHRSI(
close,
timeperiod=35,
fastk_period=12,
fastd_period=10,
fastd_matype=1)
df.loc[:, 'TRIX'] = talib.TRIX(close, timeperiod=30)
df.loc[:, 'ULTOSC'] = talib.ULTOSC(high,
low,
close,
timeperiod1=14,
timeperiod2=28,
timeperiod3=35)
df.loc[:, 'WILLR'] = talib.WILLR(high, low, close, timeperiod=35)
# overlap studies
df.loc[:,
'BBANDS_upper'], df.loc[:,
'BBANDS_middle'], df.loc[:,
'BBANDS_lower'] = talib.BBANDS(
close,
timeperiod=
12,
nbdevup=2,
nbdevdn=2,
matype=0)
df.loc[:, 'DEMA'] = talib.DEMA(close, timeperiod=30)
df.loc[:, 'EMA'] = talib.EMA(close, timeperiod=7)
df.loc[:, 'HT_TRENDLINE'] = talib.HT_TRENDLINE(close)
df.loc[:, 'KAMA'] = talib.KAMA(close, timeperiod=5)
df.loc[:, 'MA'] = talib.MA(close, timeperiod=5, matype=0)
df.loc[:, 'MIDPOINT'] = talib.MIDPOINT(close, timeperiod=20)
df.loc[:, 'WMA'] = talib.WMA(close, timeperiod=15)
df.loc[:, 'SMA'] = talib.SMA(close)
# pattern recoginition
df.loc[:, 'CDL2CROWS'] = talib.CDL2CROWS(open_price, high, low, close)
df.loc[:, 'CDL3BLACKCROWS'] = talib.CDL3BLACKCROWS(
open_price, high, low, close)
df.loc[:, 'CDL3INSIDE'] = talib.CDL3INSIDE(open_price, high, low,
close)
df.loc[:, 'CDL3LINESTRIKE'] = talib.CDL3LINESTRIKE(
open_price, high, low, close)
# price transform
df.loc[:, 'WCLPRICE'] = talib.WCLPRICE(high, low, close)
# statistic funcitons
df.loc[:, 'BETA'] = talib.BETA(high, low, timeperiod=20)
df.loc[:, 'CORREL'] = talib.CORREL(high, low, timeperiod=20)
df.loc[:, 'STDDEV'] = talib.STDDEV(close, timeperiod=20, nbdev=1)
df.loc[:, 'TSF'] = talib.TSF(close, timeperiod=20)
df.loc[:, 'VAR'] = talib.VAR(close, timeperiod=20, nbdev=1)
# volatility indicators
df.loc[:, 'ATR'] = talib.ATR(high, low, close, timeperiod=7)
df.loc[:, 'NATR'] = talib.NATR(high, low, close, timeperiod=20)
df.loc[:, 'TRANGE'] = talib.TRANGE(high, low, close)
# volume indicators
df.loc[:, 'AD'] = talib.AD(high, low, close, volume)
df.loc[:, 'ADOSC'] = talib.ADOSC(high,
low,
close,
volume,
fastperiod=10,
slowperiod=20)
df.loc[:, 'OBV'] = talib.OBV(close, volume)
# df.fillna(df.mean(), inplace=True)
df.dropna(inplace=True)
df.set_index('Date', inplace=True)
print(colored('> caching' + asset + '/' + currency + ':)', 'cyan'))
train_size = round(
len(df) *
DF_TRAIN_SIZE) # 75% to train -> test with different value
df_train = df[:train_size]
df_rollout = df[train_size:]
df_train.to_csv(df_train_path)
df_rollout.to_csv(df_rollout_path)
df_train = pd.read_csv(
df_train_path) # re-read to avoid indexing issue w/ Ray
df_rollout = pd.read_csv(df_rollout_path)
else:
print(
colored(
emoji.emojize('> ' + random.choice(emojis) + ' feching ' +
asset + '/' + currency + ' from cache',
use_aliases=True), 'magenta'))
# print(colored('> feching ' + asset + '/' + currency + ' from cache :)', 'magenta'))
df_train = pd.read_csv(df_train_path)
df_rollout = pd.read_csv(df_rollout_path)
# df_train.set_index('Date', inplace=True)
# df_rollout.set_index('Date', inplace=True)
return df_train, df_rollout
fastd_matype=0)
#STOCHRSI - Stochastic Relative Strength Index
stochrsi_fastk, stochrsi_fastd = ta.STOCHRSI(close,
timeperiod=14,
fastk_period=5,
fastd_period=3,
fastd_matype=0)
#TRIX - 1-day Rate-Of-Change (ROC) of a Triple Smooth EMA
trix = ta.TRIX(close, timeperiod=30)
#ULTOSC - Ultimate Oscillator
ultosc = ta.ULTOSC(high,
low,
close,
timeperiod1=7,
timeperiod2=14,
timeperiod3=28)
#WILLR - Williams' %R
willr = ta.WILLR(high, low, close, timeperiod=14)
df_save = pd.DataFrame(
data={
'Date': np.array(df['date']),
'adx': adx,
'adxr': adxr,
'apo': apo,
'aroondown': aroondown,
'aroonup': aroonup,
'aroonosc': aroonosc,
