def run_live_predict(trained_model, step = 1, future_target = 0):
hist_wrapper = deribit_wrapper.DeribitWrapper()
BATCH_SIZE = 1
df = hist_wrapper.fetch_chart_data(days_lookback=1)
feature_cols = ["volume", "spread_2", "spread"]
df["spread"] = df["high"] - df["low"]
df["spread_2"] = df["open"] - df["close"]
features = df[feature_cols].values
#print(features[-1])
means = features.mean(axis=0)
stds = features.std(axis=0)
dataset = (features - means) / stds
past_history = 144
start_idx = len(df) - (past_history+10)
x_val_single, y_val_single = multivariate_data(dataset, dataset[:, 2],
start_idx, None, past_history,
future_target, step,
single_step=True)
val_data_single = tf.data.Dataset.from_tensor_slices((x_val_single, y_val_single))
val_data_single = val_data_single.batch(BATCH_SIZE).repeat()
factor = 0
for x, y in val_data_single.take(1):
prediction = trained_model.predict(x)[0]
denormed = (prediction * stds[2]) + means[2]
target_denorm = (y_val_single[-1] * stds[2]) + means[2]
print(denormed, target_denorm)
factor = (target_denorm - denormed) / target_denorm
return factor
def get_predictor(STEP = 1, future_target = 5):
TRAIN_SPLIT = 6000
BUFFER_SIZE = 455
BATCH_SIZE = 255
EVALUATION_INTERVAL = 20
EPOCHS = 21
tf.random.set_seed(13)
feature_cols = ["volume", "spread_2", "spread"]
hist_wrapper = deribit_wrapper.DeribitWrapper()
df = hist_wrapper.fetch_chart_data()
df["spread"] = df["high"] - df["low"]
df["spread_2"] = df["open"] - df["close"]
features = df[feature_cols].values
means = features.mean(axis=0)
stds = features.std(axis=0)
dataset = (features - means) / stds
past_history = 144
x_train_single, y_train_single = multivariate_data(dataset, dataset[:, 2], 0,
TRAIN_SPLIT, past_history,
future_target, STEP,
single_step=True)
x_val_single, y_val_single = multivariate_data(dataset, dataset[:, 2],
TRAIN_SPLIT, None, past_history,
future_target, STEP,
single_step=True)
train_data_single = tf.data.Dataset.from_tensor_slices((x_train_single, y_train_single))
train_data_single = train_data_single.cache().shuffle(BUFFER_SIZE).batch(BATCH_SIZE).repeat()
val_data_single = tf.data.Dataset.from_tensor_slices((x_val_single, y_val_single))
val_data_single = val_data_single.batch(BATCH_SIZE).repeat()
# setup model
single_step_model = tf.keras.models.Sequential()
single_step_model.add(tf.keras.layers.LSTM(89,
input_shape=x_train_single.shape[-2:]))
single_step_model.add(tf.keras.layers.Dense(1))
single_step_model.compile(optimizer=tf.keras.optimizers.RMSprop(), loss='mae')
single_step_history = single_step_model.fit(train_data_single, epochs=EPOCHS,
steps_per_epoch=EVALUATION_INTERVAL,
validation_data=val_data_single,
validation_steps=50)
for x, y in val_data_single.take(1):
predicted = single_step_model.predict(x)[0]
plot = show_plot([x[0][:, 1].numpy(), y[0].numpy(),
predicted], 1,
'Single Step Prediction')
plot.show()
return single_step_model
def run_live_predict(trained_model, step=1, future_target=0):
hist_wrapper = deribit_wrapper.DeribitWrapper()
BATCH_SIZE = 1
df = hist_wrapper.fetch_chart_data(days_lookback=1)
feature_cols = ["volume", "spread_roc", "spread"]
df["spread"] = df["high"] - df["low"]
df["spread_roc"] = df["spread"].rolling(window=55).mean()
df.dropna(inplace=True)
print(df.iloc[-1])
features = df[feature_cols].values
#print(features[-1])
means = features.mean(axis=0)
stds = features.std(axis=0)
dataset = (features - means) / stds
past_history = 144
start_idx = len(df) - (past_history + 10)
x_val_single, y_val_single = multivariate_data(dataset,
dataset[:, 2],
start_idx,
None,
past_history,
future_target,
step,
single_step=True)
val_data_single = tf.data.Dataset.from_tensor_slices(
(x_val_single, y_val_single))
val_data_single = val_data_single.batch(BATCH_SIZE).repeat()
factor = 0
for x, y in val_data_single.take(1):
prediction = trained_model.predict(x)[0]
denormed = (prediction * stds[2]) + means[2]
target_denorm = (y_val_single[-1] * stds[2]) + means[2]
last_spread = 0.0
i = 1
while last_spread == 0.0:
last_spread = df["spread"].iloc[-i]
i += 1
factor = (last_spread - denormed) / last_spread
if factor < 0.0:
factor = factor * -1
return factor / 10