def calc_patel_TI(self, days):
""" Calculate the technical indicators from Patel et. al. for n days
in the past
Args:
days (int): Days in the past over which to
calculate relative strength index
"""
# TA.SMA has min_periods=days-1 for some reason
# Changed it to min_periods=days
self.df['MA'] = TA.SMA(self.df, days)
# Equal to my implementation
self.df['WMA'] = TA.WMA(self.df, days)
# Equal
self.df['MOM'] = TA.MOM(self.df, days)
# Had highest_high - ohlc['close']
# After changing they are equal
self.df['STOCH'] = TA.STOCH(self.df, days)
# They didn't have period when calling STOCH, instead it was used for 3
# STOCHD is actually the mean over 3 days
self.df['STOCHD'] = TA.STOCHD(self.df, days)
# They used ewm, changed it to simple rolling
# They also had ohlc['close'].diff()[1:] which resulted in returning
# one less row
# Changed min periods
self.df['RSI'] = TA.RSI(self.df, days)
# TODO: What do they mean in the paper
# Changed min periods
self.df['MACD'] = TA.MACD(self.df, signal=days)['MACD']
self.df['WILLIAMS'] = TA.WILLIAMS(self.df, days)
self.df['ADL'] = TA.ADL(self.df)
self.df['CCI'] = TA.CCI(self.df, days)
# Drop columns we no longer need
self.df.drop(['open', 'high', 'low', 'volume'], axis=1, inplace=True)
# Drop rows with nan
self.df.dropna(inplace=True)
def get_adl(data):
"""Calculate the accumulation/distribution line of given dataframe.
:param data: a dataframe in OHLC format
:return: a Pandas series
"""
if data is None:
raise EmptyDataError("[!] Invalid data value")
result = TA.ADL(data)
if result is None:
raise IndicatorException
return result
def test_adl():
adl = TA.ADL(ohlc)
assert isinstance(adl, series.Series)
assert adl.values[-1] == 1221072523.7384958
def test_adl():
adl = TA.ADL(ohlc).round(decimals=8)
assert isinstance(adl, series.Series)
assert adl.values[-1] == 303320.96403697
# Sell when the short crosses below the medium and really sell when it crosses below the long
# So we have a number of things to implement (signal is +1 when short above medium, +2 short above long, +3 when
# medium above long as well. Similarly in the opposite direction. Positive is buy, negative is sell)
EMAV_conditions = [
(prices["Exp_mav_sh"] > prices["Exp_mav_med"]) & (prices["Exp_mav_sh"] < prices["Exp_mav_lon"]),
(prices["Exp_mav_sh"] > prices["Exp_mav_med"]) & (prices["Exp_mav_sh"] > prices["Exp_mav_lon"]),
(prices["Exp_mav_sh"] > prices["Exp_mav_lon"]) & (prices["Exp_mav_med"] > prices["Exp_mav_lon"]),
(prices["Exp_mav_sh"] < prices["Exp_mav_med"]) & (prices["Exp_mav_sh"] < prices["Exp_mav_lon"]),
(prices["Exp_mav_sh"] < prices["Exp_mav_med"]) & (prices["Exp_mav_sh"] < prices["Exp_mav_lon"]),
(prices["Exp_mav_sh"] < prices["Exp_mav_lon"]) & (prices["Exp_mav_med"] < prices["Exp_mav_lon"]),
]
EMAV_choices = [1.0, 2.0, 3.0, -1.0, -2.0, -3.0]
prices["Exp_mav-signal"] = np.select(EMAV_conditions, EMAV_choices, default=0.0)
# Accumulation and Distribution - Marc Chaikin to determine flow in or out of a security
prices["Money Flow Multiplier"] = TA.ADL(ohlcv)
# MACD oscillator
prices["MACDb-finta"] = TA.CHAIKIN(ohlcv)
prices["MFI"] = TA.MFI(ohlc)
prices["CMFV"] = prices['Volume'] * ((prices['Adj Close'] - prices['Adj Low']) -
(prices['Adj High']-prices['Adj Close'])) / \
(prices['Adj High'] - prices['Adj Low'])
# It shows how strong the trend is, if the price is rising but the indicator is falling, this indicates that
# buying or accumulation volume many not be enough to support the price rise and a decline may be forthcoming
prices["CMFV-rolling"] = prices["CMFV"].rolling(window=20).sum()
prices["CMFV-signal"] = prices["CMFV-rolling"] - prices["CMFV-rolling"].shift(periods=5)
# Moving average convergence divergence
# A divergence is where the price doesnt agree with the indicator
def test_adl():
adl = TA.ADL(ohlc)
assert isinstance(adl, series.Series)
assert adl.values[-1] == 303320.96403697244
def technical_indicators(tiDF, usdflag):
# tiDF 'date_time', 'Date', 'Time', 'Open', 'High', 'Low', 'Close', 'Volume', 'Stock'
ohlcv = tiDF[[
'date_time', 'Date', 'Open', 'High', 'Low', 'Close', 'Volume'
]]
for i in ohlcv.columns:
ohlcv = ohlcv.rename(columns={str(i): str(i.lower())})
ohlcv.rename(columns={'date_time': 'Date', 'date': 'day'}, inplace=True)
ohlcv.set_index('Date', inplace=True)
ohlcv['sma2'] = TA.SMA(ohlcv, 2)
ohlcv['sma3'] = TA.SMA(ohlcv, 3)
ohlcv['sma4'] = TA.SMA(ohlcv, 4)
ohlcv['sma5'] = TA.SMA(ohlcv, 5)
ohlcv['sma6'] = TA.SMA(ohlcv, 6)
ohlcv['sma7'] = TA.SMA(ohlcv, 7)
ohlcv['rsi'] = TA.RSI(ohlcv)
ohlcv['cci'] = TA.CCI(ohlcv)
ohlcv['adx'] = TA.ADX(ohlcv)
ohlcv['vpt'] = TA.VPT(ohlcv)
ohlcv['efi'] = TA.EFI(ohlcv)
ohlcv['wobv'] = TA.WOBV(ohlcv)
ohlcv['vzo'] = TA.VZO(ohlcv)
ohlcv['pzo'] = TA.PZO(ohlcv)
ohlcv['tp'] = TA.TP(ohlcv)
ohlcv['adl'] = TA.ADL(ohlcv)
ohlcv['smma'] = TA.SMMA(ohlcv)
ohlcv['tr'] = TA.TR(ohlcv)
ohlcv['sar'] = TA.SAR(ohlcv)
ohlcv['vwap'] = TA.VWAP(ohlcv)
ohlcv['ssma'] = TA.SSMA(ohlcv)
ohlcv['dema'] = TA.DEMA(ohlcv)
ohlcv['tema'] = TA.TEMA(ohlcv)
ohlcv['trix'] = TA.TRIX(ohlcv)
stock = tiDF['Stock'].unique().tolist()[0]
ohlcv['Stock'] = stock
ohlcv = ohlcv.dropna()
numTimes = ohlcv.groupby('day').count()
shortDays = numTimes[numTimes.sma7 != 7].index.values
ohlcv = ohlcv[~ohlcv['day'].isin(shortDays)]
ohlcv = ohlcv.drop(['high', 'low', 'close', 'day'], axis=1)
ohlcv.reset_index(level=0, inplace=True)
for i in ohlcv.columns:
ohlcv = ohlcv.rename(columns={str(i): str(i.capitalize())})
ohlcv.rename(columns={'Date': 'date_time'}, inplace=True)
pd.options.display.max_columns = None
pd.options.display.max_rows = None
if usdflag:
totale = pd.read_csv("total.csv")
totale['date_time'] = to_datetime(totale['date_time'],
format="%Y-%m-%d %H:%M:%S")
result = pd.merge(ohlcv, totale, how='inner', on=['date_time'])
else:
result = ohlcv.copy()
print(result.info())
print(result.head(5))
return result
