def preprocess(self, ts, train_percentage):
if len(ts.shape) == 1:
ts = ts.reshape(1, len(ts))
# Regress original ts
X_orig, y_orig = sliding_window(ts, p=self.q, drop_last_dim=True)
self.regressor_model.preprocess(ts, 1.0)
self.regressor_model.fit(X_orig, y_orig)
# Predict with regressor model and substract original to obtain residuals.
pred = self.regressor_model.model.predict(X_orig)
pred = pred.reshape(1, len(pred))
residual_ts = np.concatenate((np.zeros(
(1, self.q)), pred), axis=1) - ts
# Define sliding window on residuals and compute historic vols from it
X_residual, y_residual = sliding_window(residual_ts,
p=self.q,
drop_last_dim=False)
y_sigmas = X_residual.std(axis=1)
y_sigmas_squared = y_sigmas * y_sigmas
X_residual_squared = X_residual * X_residual
# Compute train-test-split and set internal variables
(X_train, y_train), (X_test, y_test) = train_test_split(
X_residual_squared,
y_sigmas_squared,
train_percentage=train_percentage)
self.X_train = X_train
self.y_train = y_train
self.X_test = X_test
self.y_test = y_test
self.has_data = True
def sliding_window_test():
X, y = sliding_window(X_dummy, p=3)
assert X.shape == (6, 3, 1)
assert y.shape == (2 * (6 - 3), )
assert np.array_equal(X[0, :, 0], np.array([1, 2, 3]))
assert np.array_equal(y[0], 4)
assert np.array_equal(X[5, :, 0], np.array([9, 10, 11]))
assert np.array_equal(y[4], 11)
def preprocess(self, ts, train_percentage):
if len(ts.shape) == 1:
ts = ts.reshape(1, len(ts))
X, y = sliding_window(ts, p=self.p)
(X_train, y_train), (X_test, y_test) = train_test_split(
X, y, train_percentage=train_percentage)
self.X_train = X_train
self.y_train = y_train
self.X_test = X_test
self.y_test = y_test
self.has_data = True
def sliding_window_multi_test():
X, y = sliding_window(X_mult_dummy, p=3)
assert X.shape == (3, 3, 2)
assert y.shape == (6 - 3, 2)