def build_ta_feature(symbol: str, daily: Sequence[tuple],
save_feature: bool = True) -> pd.DataFrame:
'''Build features using the daily time series data. Indicators like EMA
should be computed using the complete set of data (instead of a subset)
@param: symbol: stock symbol that the daily data belongs to
@param: daily: daily time series data for a symbol
@param: save_feature: whether to save the features to file
'''
logger.info('Build ta feature for: %s. Last daily date: %s', symbol, daily[-1][0])
col_dtype = {'date': object, 'open': 'float64', 'high': 'float64', 'low': 'float64',
'close': 'float64', 'adj_close': 'float64', 'volume': 'int32'}
df = pd.DataFrame(daily, columns=col_dtype.keys(), dtype=object)
df = df.astype(col_dtype, copy=True)
high, low, close, volume = df['high'], df['low'], df['close'], df['volume']
# Overlap Studies
upperband, middleband, lowerband = Indicator.BBANDS(close, timeperiod=5)
df['ub_r'], df['mb_r'], df['lb_r'] = upperband / close, middleband / close, lowerband / close
df['trima_r'] = Indicator.TRIMA(close, timeperiod=30) / close
df['wma_r'] = Indicator.WMA(close, timeperiod=30) / close
# Momentum Indicator
macd, macdsignal, macdhist = Indicator.MACD(close, fastperiod=12, slowperiod=26, signalperiod=9)
df['macdhist'] = macdhist
df['ppo'] = Indicator.PPO(close, fastperiod=12, slowperiod=26)
df['rsi'] = Indicator.RSI(close, timeperiod=14)
# Volume Indicator
df['adosc'] = Indicator.ADOSC(high, low, close, volume * 1.0, fastperiod=3, slowperiod=10)
# Volatility Indicator
df['natr'] = Indicator.NATR(high, low, close, timeperiod=14)
# Drop rows with NaN values
df.dropna(axis=0, how='any', inplace=True)
# Drop useless columns
col_drop = ['open', 'high', 'low', 'close', 'volume']
df.drop(columns=col_drop, inplace=True)
if save_feature:
df.to_csv(_data_folder + '{0}_ta.csv'.format(symbol), float_format='%.4f', index=False)
return df
temp = name.split('-')
temp_str = temp[0] + temp[1]
model_name = temp_str.lower().split('.')[0]
BASELINE_MODEL_PATH = 'model/model-baseline-new/model-{}'.format(
model_name)
PSO_MODEL_PATH = 'model/model-pso-new/model-pso-{}'.format(model_name)
data = pd.read_csv(FILE_PATH)
ann = keras.models.load_model(BASELINE_MODEL_PATH)
pso = keras.models.load_model(PSO_MODEL_PATH)
# iteration
print('starting for {}'.format(name.split('.')[0]))
# create indicator
indicator = Indicator(data)
indicator_data = indicator.RSI()
indicator_data = indicator.EMA()
indicator_data = indicator.MACD()
indicator_data.dropna(inplace=True)
indicator_data['change_of_ema'] = (
(indicator_data['Close'] - indicator_data['ema_5_day']) /
indicator_data['ema_5_day']) * 100
data_set = indicator_data[[
'Close', 'rsi', 'Histogram', 'change_of_ema', 'change'
]]
simulator = Simulator(simulate_day=N_DAY,
simulate_data=data_set,
baseline_model=ann,
pso_model=pso)
simulator.start()