BATCH_SIZE = 128
BUFFER_SIZE = 10000
EPOCHS = 10
CLASSIFICATION = True
step = 1
history_size = 21
target_distance = 2
features_considered = [
'Close', 'Volume', 'MA_short', 'MA_long', 'Wilders_EMA', 'bb_bbm',
'bb_bbh', 'bb_bbl', 'bb_bbhi', 'bb_bbli'
]
# features_considered = ['Close', 'MA_short', 'MA_long', 'Wilders_EMA']
# Initialize Bollinger Bands Indicator
e = EquityData('data/SPY.csv', 'SPY')
indicator_bb = BollingerBands(close=e.close(), n=20, ndev=2)
e.data['MA_short'] = moving_average(e, window=history_size)
e.data['MA_long'] = moving_average(e, window=5)
e.data['Wilders_EMA'] = e.close().ewm(alpha=1 / history_size,
adjust=False).mean()
# Add Bollinger Bands features
e.data['bb_bbm'] = indicator_bb.bollinger_mavg()
e.data['bb_bbh'] = indicator_bb.bollinger_hband()
e.data['bb_bbl'] = indicator_bb.bollinger_lband()
# Add Bollinger Band high indicator
e.data['bb_bbhi'] = indicator_bb.bollinger_hband_indicator()
# Add Bollinger Band low indicator
e.data['bb_bbli'] = indicator_bb.bollinger_lband_indicator()
e.data = e.data[21:]
import tensorflow as tf
from utils import plot_train_history, multi_step_plot
from technical_analysis import moving_average
BATCH_SIZE = 128
BUFFER_SIZE = 10000
tf.random.set_seed(42)
step = 1
history_size = 30
target_distance = 1
features_considered = ['Close', 'Volume', 'MA_short', 'MA_long']
e = EquityData('data/SPY.csv', 'SPY')
e.data['MA_short'] = moving_average(e, window=21)
e.data['MA_long'] = moving_average(e, window=5)
e.data = e.data[21:]
EVALUATION_INTERVAL = int(e.data.shape[0]/BATCH_SIZE) * 1
features = e.data[features_considered]
assert(list(features)[0] == 'Close')
features.index = e.date()
dataset = features.values
x_train_multi, y_train_multi, x_val_multi, y_val_multi = split_multivariate(dataset, history_size, target_distance, step, single_step=False)
print ('Single window of past history : {}'.format(x_train_multi[0].shape))
print ('\n Target temperature to predict : {}'.format(y_train_multi[0].shape))
from market import EquityData
from models.lstm import split, show_plot
import tensorflow as tf
tf.random.set_seed(42)
BATCH_SIZE = 256
BUFFER_SIZE = 10000
e = EquityData('data/SPY.csv', 'SPY')
x_train_uni, y_train_uni, x_val_uni, y_val_uni = split(e)
train_univariate = tf.data.Dataset.from_tensor_slices(
(x_train_uni, y_train_uni))
train_univariate = train_univariate.cache().shuffle(BUFFER_SIZE).batch(
BATCH_SIZE).repeat()
val_univariate = tf.data.Dataset.from_tensor_slices((x_val_uni, y_val_uni))
val_univariate = val_univariate.batch(BATCH_SIZE).repeat()
simple_lstm_model = tf.keras.models.Sequential([
tf.keras.layers.LSTM(8, input_shape=x_train_uni.shape[-2:]),
tf.keras.layers.Dense(1)
])
simple_lstm_model.compile(optimizer='adam', loss='mae')
for x, y in val_univariate.take(1):
print(simple_lstm_model.predict(x).shape)
EVALUATION_INTERVAL = 200
from market import EquityData
from models.lstm import split_multivariate
from utils import create_time_steps
import matplotlib.pyplot as plt
import tensorflow as tf
import numpy as np
features_considered = ['Close', 'Volume']
e = EquityData('data/SPY.csv', 'SPY')
features = e.data[features_considered]
features.index = e.date()
step = 1
history_size = 90
target_distance = 14
dataset = features.values
train_split = int(len(dataset) * 0.7)
data_mean = dataset[:train_split].mean(axis=0)
data_std = dataset[:train_split].std(axis=0)
data_no_transform = dataset[-history_size:]
data_no_transform = np.reshape(
data_no_transform,
(1, data_no_transform.shape[0], data_no_transform.shape[1]))
dataset = (dataset - data_mean) / data_std
data = dataset[-history_size:]
data = np.reshape(data, (1, data.shape[0], data.shape[1]))
# model = tf.keras.models.load_model('saved_models/multivariate_multi_model')
ssm.add(tf.keras.layers.LSTM(128))
ssm.add(tf.keras.layers.Dense(64))
ssm.add(tf.keras.layers.Dense(1, activation='sigmoid'))
ssm.compile(optimizer=tf.keras.optimizers.Adam(lr=0.001),
loss=tf.keras.losses.BinaryCrossentropy(),
metrics=['accuracy'])
return ssm
if __name__ == "__main__":
# construct data and get additional info
e = EquityData(DATA_DIR, DATA_SYM)
data = pyfinancialdata.get_multi_year(provider='histdata',
instrument='SPXUSD',
years=[2016, 2017, 2018],
time_group='10min')
e.data = data
# add MACD
e.data['MA_long'] = e.data['close'].rolling(window=52).mean()
e.data['MA_short'] = e.data['close'].rolling(window=7).mean()
# evaluation interval
window = int(e.data.shape[0] / BATCH_SIZE) * 1
# pick selected features
features = e.data[FEATURES]
import tensorflow as tf
from utils import plot_train_history, multi_step_plot
from technical_analysis import moving_average
tf.random.set_seed(42)
BATCH_SIZE = 128
BUFFER_SIZE = 10000
EPOCHS = 15
step = 1
history_size = 30
target_distance = 1
features_considered = ['Close', 'Volume', 'MA_short', 'MA_long', 'Wilders_EMA']
e = EquityData('data/SPY.csv', 'SPY')
e.data['MA_short'] = moving_average(e, window=21)
e.data['MA_long'] = moving_average(e, window=5)
e.data['Wilders_EMA'] = e.close().ewm(alpha=1/15, adjust=False).mean()
e.data = e.data[21:]
EVALUATION_INTERVAL = int(e.data.shape[0]/BATCH_SIZE) * 1
features = e.data[features_considered]
assert(list(features)[0] == 'Close')
features.index = e.date()
features.plot(subplots=True)
plt.show()
dataset = features.values
x_train_multi, y_train_multi, x_val_multi, y_val_multi = split_multivariate(dataset, history_size, target_distance, step, single_step=False)
history_size = 180
target_distance = 6
features_considered = [
'Close', 'Volume', 'MA_short', 'MA_long', 'Wilders_EMA', 'bb_bbm',
'bb_bbh', 'bb_bbl', 'bb_bbhi', 'bb_bbli'
]
features_considered = [
'Close', 'Volume', 'Wilders_EMA', 'bb_bbm', 'bb_bbh', 'bb_bbl', 'bb_bbhi',
'bb_bbli', 'trend_adx', 'momentum_stoch_signal', 'trend_trix',
'trend_ichimoku_a', 'momentum_kama', 'momentum_tsi', 'volatility_kcli'
]
features_considered = ['close', 'MA_long', 'MA_short']
# features_considered = ['Close', 'MA_short', 'MA_long', 'Wilders_EMA']
# Initialize Bollinger Bands Indicator
e = EquityData('data/SPY.csv', 'SPY')
data = pyfinancialdata.get_multi_year(provider='histdata',
instrument='SPXUSD',
years=[2016, 2017, 2018],
time_group='10min')
e.data = data
indicator_bb = BollingerBands(close=e.data['close'], n=20, ndev=2)
# e.data = add_all_ta_features(e.data, open="open", high="high", low="low", close="close")
e.data['MA_long'] = e.data['close'].rolling(window=52).mean()
e.data['MA_short'] = e.data['close'].rolling(window=7).mean()
# e.data['MA_short'] = moving_average(e, window=5)
# e.data['Wilders_EMA'] = e.close().ewm(alpha=1/history_size, adjust=False).mean()
# # Add Bollinger Bands features
# e.data['bb_bbm'] = indicator_bb.bollinger_mavg()
def test_equity_loads_data():
e = EquityData('data/SPY.csv', 'SPY')
assert (type(e.data) == dict)