def ta_volume_as_time(df: _t.PatchedPandas, volume="Volume"):
if df.ndim > 1:
res = df.copy()
res.index = df[volume].cumsum()
return res
else:
return _pd.Series(df.index, index=df.cumsum())
def ta_z_norm(df: Typing.PatchedPandas,
period=200,
ddof=1,
demean=True,
lag=0):
# (value - mean) / std
s = df.rolling(period).std(ddof=ddof)
a = (df - df.rolling(period).mean().shift(lag)) if demean else df
return (a / s / 4).rename(df.name)
def ta_z_norm(df: Typing.PatchedPandas, period=200, ddof=1, demean=True, lag=0):
if has_indexed_columns(df):
return pd.DataFrame(
{col: ta_z_norm(df[col], period, ddof, demean) for col in df.columns},
index=df.index
)
# (value - mean) / std
s = df.rolling(period).std(ddof=ddof)
a = (df - df.rolling(period).mean().shift(lag)) if demean else df
return (a / s / 4).rename(df.name)
def ta_ncdf_compress(df: Typing.PatchedPandas, period=200, lower_percentile=25, upper_percentile=75) -> Typing.PatchedPandas:
if has_indexed_columns(df):
return pd.DataFrame(
{col: ta_ncdf_compress(df[col], period, lower_percentile, upper_percentile) for col in df.columns},
index=df.index
)
f50 = df.rolling(period).mean().rename("f50")
fup = df.rolling(period).apply(lambda x: np.percentile(x, upper_percentile)).rename("fup")
flo = df.rolling(period).apply(lambda x: np.percentile(x, lower_percentile)).rename("flo")
return pd.Series(norm.cdf(0.5 * (df - f50) / (fup - flo)) - 0.5, index=df.index, name=df.name)
def ta_ewma_covariance(df: Typing.PatchedPandas,
convert_to='returns',
alpha=0.97):
data = df.copy()
if convert_to == 'returns':
data = df.pct_change()
if convert_to == 'log-returns':
data = _np.log(df) - _np.log(df.shift(1))
data.columns = data.columns.to_list()
return data.ewm(com=alpha).cov()
def ta_ncdf_compress(df: Typing.PatchedPandas,
period=200,
lower_percentile=25,
upper_percentile=75) -> Typing.PatchedPandas:
f50 = df.rolling(period).mean().rename("f50")
fup = df.rolling(period).apply(
lambda x: np.percentile(x, upper_percentile)).rename("fup")
flo = df.rolling(period).apply(
lambda x: np.percentile(x, lower_percentile)).rename("flo")
return pd.Series(norm.cdf(0.5 * (df - f50) / (fup - flo)) - 0.5,
index=df.index,
name=df.name)
def ta_backtest(signal: Typing.PatchedDataFrame,
prices: Typing.PatchedPandas,
action: Callable[[pd.Series], Tuple[int, float]],
slippage: Callable[[float], float] = lambda _: 0):
if has_indexed_columns(signal):
assert len(signal.columns) == len(
prices.columns), "Signal and Prices need the same shape!"
res = pd.DataFrame({},
index=signal.index,
columns=pd.MultiIndex.from_product([[], []]))
for i in range(len(signal.columns)):
df = ta_backtest(signal[signal.columns[i]],
prices[prices.columns[i]], action, slippage)
top_level_name = ",".join(prices.columns[i]) if isinstance(
prices.columns[i], tuple) else prices.columns[i]
df.columns = pd.MultiIndex.from_product([[top_level_name],
df.columns.to_list()])
res = res.join(df)
return res
assert isinstance(prices, pd.Series), "prices need to be a series!"
trades = StreamingTransactionLog()
def trade_log_action(row):
direction_amount = action(row)
if isinstance(direction_amount, tuple):
trades.perform_action(*direction_amount)
else:
trades.rebalance(float(direction_amount))
signal.to_frame().apply(trade_log_action, axis=1, raw=True)
return trades.evaluate(prices.rename("price"), slippage)
def ta_week(po: _t.PatchedPandas):
if not isinstance(po.index, _pd.DatetimeIndex):
df = po.copy()
df.index = _pd.to_datetime(df.index)
else:
df = po
return (df.index.week / 52.0).to_series(index=po.index, name="week")
def ta_sinusoidial_week_day(po: _t.PatchedPandas):
if not isinstance(po.index, _pd.DatetimeIndex):
df = po.copy()
df.index = _pd.to_datetime(df.index)
else:
df = po
return _np.sin(2 * _np.pi * (df.index.dayofweek / 6.0)).to_series(
index=po.index, name="dow")
def ta_sinusoidal_week(po: _t.PatchedPandas):
if not isinstance(po.index, _pd.DatetimeIndex):
df = po.copy()
df.index = _pd.to_datetime(df.index)
else:
df = po
return _np.sin(2 * _np.pi *
(df.index.isocalendar().week / 52.0)).rename("week")
def ta_inverse_gasf(df: Typing.PatchedPandas):
if has_indexed_columns(df):
# todo implement recursion
pass
def gaf_decode(x):
values = x._.values
values = values.reshape(values.shape[1:])
return _np.sqrt(((_np.diag(values) + 1) / 2)).tolist()
return df.to_frame().apply(gaf_decode, axis=1, result_type='expand')
def ta_mean_returns(df: Typing.PatchedPandas, period=20) -> _PANDAS:
return _wcs(f"mean_return_{period}",
df.pct_change().rolling(period).mean())
def ta_performance(df: Typing.PatchedPandas):
delta = df.pct_change() + 1
return delta.cumprod()
def ta_log_returns(df: Typing.PatchedPandas, period=1):
current = df
lagged = df.shift(period)
return _wcs("log_return", np.log(current) - np.log(lagged))
def ta_returns(df: Typing.PatchedPandas, period=1):
return _wcs("return", df.pct_change(periods=period))
def ta_returns(df: Typing.PatchedPandas):
return _wcs("return", df.pct_change())
