def _feature_engineering(df, volume=True):
if volume:
df = ta.add_all_ta_features(df,
open="Open",
high="High",
low="Low",
close="Last",
volume="Volume",
fillna=False)
else:
df = ta.add_momentum_ta(df,
high="High",
low="Low",
close="Last",
volume="Volume",
fillna=False)
df = ta.add_volatility_ta(df,
high="High",
low="Low",
close="Last",
fillna=False)
df = ta.add_trend_ta(df,
high="High",
low="Low",
close="Last",
fillna=False)
df = ta.add_others_ta(df, close="Last", fillna=False)
df["trend_psar_up"] = df["trend_psar_up"].fillna(0.0)
df["trend_psar_down"] = df["trend_psar_down"].fillna(0.0)
return df
def add_technical_indicators_others(stock_dataframe: pd.DataFrame,
open="Open",
high="High",
low="Low",
close="Close",
volume="Volume") -> pd.DataFrame:
return add_others_ta(df=stock_dataframe, close=close)
def get_others_indicators(df, threshold=0.5, plot=False):
df_others = df.copy()
# add custom indicators
# ...
df_others = add_others_ta(df_others, close="Close")
return DropCorrelatedFeatures(df_others, threshold, plot)
def get_features(df, options=None):
if options == 'all':
df = ta.add_all_ta_features(df, "open", "high", "low", "close",
"volume")
elif options == 'volume':
df = ta.add_volume_ta(df, "high", "low", "close", "volume")
elif options == 'momentum':
df = ta.add_momentum_ta(df, "high", "low", "close", "volume")
elif options == 'volatility':
df = ta.add_volatility_ta(df, "high", "low", "close")
elif options == 'trend':
df = ta.add_trend_ta(df, "high", "low", "close")
elif options == 'others':
df = ta.add_others_ta(df, "close")
df = df.sort_values('date')
return df
def add_custom_ta_features(
df: pd.DataFrame,
open: str, # noqa
high: str,
low: str,
close: str,
fillna: bool = False,
colprefix: str = "",
apply_pct: bool = False,
plot: bool = False,
) -> pd.DataFrame:
# Add Volatility TA
df = ta.add_volatility_ta(df=df,
high=high,
low=low,
close=close,
fillna=fillna,
colprefix=colprefix)
# Add Trend TA
df = ta.add_trend_ta(df=df,
high=high,
low=low,
close=close,
fillna=fillna,
colprefix=colprefix)
# Add Other TA
df = ta.add_others_ta(df=df,
close=close,
fillna=fillna,
colprefix=colprefix)
# convert to pct
if apply_pct:
df = df.pct_change(fill_method='ffill')
df = df.applymap(lambda x: x * 100)
df.replace([np.inf, -np.inf], np.nan, inplace=True)
df.astype(np.float32)
df = df.round(5)
if fillna:
df.fillna(value=0, inplace=True)
if plot:
fig = make_subplots(rows=5,
cols=1,
shared_xaxes=True,
vertical_spacing=0.02,
subplot_titles=("Close", "Bollinger Bands",
"MACD"))
fig.add_trace(go.Scatter(x=df.index, y=df['close'], name=symbol),
row=1,
col=1)
# Bollinger-Bands
fig.add_trace(go.Scatter(x=df.index, y=df['close'], name=symbol),
row=2,
col=1)
fig.add_trace(go.Scatter(x=df.index,
y=df['volatility_bbh'],
name=symbol + ' High BB'),
row=2,
col=1)
fig.add_trace(go.Scatter(x=df.index,
y=df['volatility_bbl'],
name=symbol + ' Low BB'),
row=2,
col=1)
fig.add_trace(go.Scatter(x=df.index,
y=df['volatility_bbm'],
name=symbol + ' EMA BB'),
row=2,
col=1)
# MACD
fig.add_trace(go.Scatter(x=df.index,
y=df['trend_macd'],
name=symbol + ' MACD'),
row=3,
col=1)
fig.add_trace(go.Scatter(x=df.index,
y=df['trend_macd_signal'],
name=symbol + ' MACD Signal'),
row=3,
col=1)
fig.add_trace(go.Scatter(x=df.index,
y=df['trend_macd_diff'],
name=symbol + ' MACD Difference'),
row=3,
col=1)
# SMA
fig.add_trace(go.Scatter(x=df.index, y=df['close'], name=symbol),
row=4,
col=1)
fig.add_trace(go.Scatter(x=df.index,
y=df['trend_sma_fast'],
name=symbol + ' SMA-Fast'),
row=4,
col=1)
fig.add_trace(go.Scatter(x=df.index,
y=df['trend_sma_slow'],
name=symbol + ' SMA-Slow'),
row=4,
col=1)
# EMA
fig.add_trace(go.Scatter(x=df.index, y=df['close'], name=symbol),
row=5,
col=1)
fig.add_trace(go.Scatter(x=df.index,
y=df['trend_ema_fast'],
name=symbol + ' EMA-Fast'),
row=5,
col=1)
fig.add_trace(go.Scatter(x=df.index,
y=df['trend_ema_slow'],
name=symbol + ' EMA-Slow'),
row=5,
col=1)
config = {'displayModeBar': False}
fig.show(config=config)
return df
def transform(self, X, y=None) -> pd.DataFrame:
"""
Calculates features (technical indicators) of selected type.
Parameters
----------
X : iterable
Data in OHLCV format.
y : iterable, default=None
Training targets.
Returns
-------
Xt : array-like of shape (n_samples, n_transformed_features)
Dataset with calculated features.
"""
if self.all_features:
return add_all_ta_features(
df=X,
open="open",
high="high",
low="low",
close="close",
volume="volume",
fillna=self.fillna,
colprefix=self.colprefix,
)
if self.volume_features:
X = add_volume_ta(
df=X,
high="high",
low="low",
close="close",
volume="volume",
fillna=self.fillna,
colprefix=self.colprefix,
)
if self.volatility_features:
X = add_volatility_ta(
df=X,
high="high",
low="low",
close="close",
fillna=self.fillna,
colprefix=self.colprefix,
)
if self.trend_features:
X = add_trend_ta(
df=X,
high="high",
low="low",
close="close",
fillna=self.fillna,
colprefix=self.colprefix,
)
if self.momentum_features:
X = add_momentum_ta(
df=X,
high="high",
low="low",
close="close",
volume="volume",
fillna=self.fillna,
colprefix=self.colprefix,
)
if self.others_features:
X = add_others_ta(df=X,
close="close",
fillna=self.fillna,
colprefix=self.colprefix)
return X
def makePlot(type):
# Fetch the data
# Option 1: Use CryptoDataDownload
#cdd = CryptoDataDownload()
#df = cdd.fetch("Bitfinex", "USD", "BTC", "d")
# Option 2: Use Binance data
# Found here: https://github.com/StephanAkkerman/TensorTrade
coin = "BAT"
df = fetchData(symbol=(coin + "USDT"), amount=2, timeframe='4h')
# Drop unix and set 'date' as index
df = df[['date', 'open', 'high', 'low', 'close', 'volume']]
df = df.set_index("date")
# Calculate absolute correlation and clean up names
if (type == 'momentum'):
df = ta.add_momentum_ta(df,
high='high',
low='low',
close='close',
volume='volume').corr().abs()
df.columns = df.columns.str.replace('momentum_', '')
df.index = df.index.str.replace('momentum_', '')
if (type == 'volume'):
df = ta.add_volume_ta(df,
high='high',
low='low',
close='close',
volume='volume').corr().abs()
df.columns = df.columns.str.replace('volume_', '')
df.index = df.index.str.replace('volume_', '')
if (type == 'trend'):
df = ta.add_trend_ta(df, high='high', low='low',
close='close').corr().abs()
df.columns = df.columns.str.replace('trend_', '')
df.index = df.index.str.replace('trend_', '')
if (type == 'volatility'):
df = ta.add_volatility_ta(df, high='high', low='low',
close='close').corr().abs()
df.columns = df.columns.str.replace('volatility_', '')
df.index = df.index.str.replace('volatility_', '')
if (type == 'others'):
df = ta.add_others_ta(df, close='close').corr().abs()
df.columns = df.columns.str.replace('volumeothers_', '')
df.index = df.index.str.replace('volumeothers_', '')
if (type == 'all'):
df = ta.add_all_ta_features(df,
open='open',
high='high',
low='low',
close='close',
volume='volume').corr().abs()
# Remove this from heatmap
df = df.drop(index=['open', 'high', 'low', 'close', 'volume'],
columns=['open', 'high', 'low', 'close', 'volume'])
# Plot the heatmap
plt.subplots(figsize=(10, 10))
# https://seaborn.pydata.org/tutorial/color_palettes.html for possible colors
sns.heatmap(df,
annot=False,
linewidth=0,
center=0.5,
cmap=sns.color_palette("viridis", as_cmap=True))
plt.tight_layout()
plt.yticks(rotation=0)
plt.show()