def test_n_info_values():
assert_raises(
ValueError, FROLS, n_info_values=1.2, basis_function=Polynomial(degree=2)
)
assert_raises(
ValueError, FROLS, n_info_values=-1, basis_function=Polynomial(degree=2)
)
def test_fit_polynomial():
basis_function = Polynomial(degree=2)
data = np.array(([1, 1, 1], [2, 3, 4], [3, 3, 3]))
max_lag = 1
output = np.array([[4, 6, 8, 9, 12, 16], [9, 9, 9, 9, 9, 9]])
r = basis_function.fit(data=data, max_lag=max_lag)
assert_array_equal(output, r)
def test_model_order_selection():
assert_raises(
TypeError, FROLS, order_selection=1, basis_function=Polynomial(degree=2)
)
assert_raises(
TypeError, FROLS, order_selection="True", basis_function=Polynomial(degree=2)
)
assert_raises(
TypeError, FROLS, order_selection=None, basis_function=Polynomial(degree=2)
)
def test_fit_polynomial_predefined():
basis_function = Polynomial(degree=2)
data = np.array(([1, 1, 1], [2, 3, 4], [3, 3, 3]))
max_lag = 1
predefined_regressors = np.array([0, 2, 4])
output = np.array([[4, 8, 12], [9, 9, 9]])
r = basis_function.fit(
data=data, max_lag=max_lag, predefined_regressors=predefined_regressors
)
assert_array_equal(output, r)
def test_estimate_parameter_error():
assert_raises(
TypeError,
SimulateNARMAX,
estimate_parameter=1,
basis_function=Polynomial(degree=2),
)
def test_model_type_error():
assert_raises(
ValueError,
SimulateNARMAX,
model_type="NFAR",
basis_function=Polynomial(degree=2),
)
def test_model_prediction():
x, y, theta = create_test_data()
basis_function = Polynomial(degree=2)
train_percentage = 90
split_data = int(len(x) * (train_percentage / 100))
X_train = x[0:split_data, 0]
X_test = x[split_data::, 0]
y1 = y[0:split_data, 0]
y_test = y[split_data::, 0]
y_train = y1.copy()
y_train = np.reshape(y_train, (len(y_train), 1))
X_train = np.reshape(X_train, (len(X_train), 1))
y_test = np.reshape(y_test, (len(y_test), 1))
X_test = np.reshape(X_test, (len(X_test), 1))
model = FROLS(
n_terms=5,
extended_least_squares=False,
ylag=[1, 2],
xlag=2,
estimator="least_squares",
basis_function=basis_function,
)
model.fit(X=X_train, y=y_train)
assert_raises(Exception, model.predict, X=X_test, y=y_test[:1])
def test_extended_least_squares():
assert_raises(
TypeError, FROLS, extended_least_squares=1, basis_function=Polynomial(degree=2)
)
assert_raises(
TypeError,
FROLS,
extended_least_squares="True",
basis_function=Polynomial(degree=2),
)
assert_raises(
TypeError,
FROLS,
extended_least_squares=None,
basis_function=Polynomial(degree=2),
)
def test_default_values():
default = {
"estimator": "recursive_least_squares",
"extended_least_squares": False,
"lam": 0.98,
"delta": 0.01,
"offset_covariance": 0.2,
"mu": 0.01,
"eps": np.finfo(np.float64).eps,
"gama": 0.2,
"weight": 0.02,
"model_type": "NARMAX",
"estimate_parameter": True,
"calculate_err": False,
}
model = SimulateNARMAX(basis_function=Polynomial())
model_values = [
model.estimator,
model._extended_least_squares,
model._lam,
model._delta,
model._offset_covariance,
model._mu,
model._eps,
model._gama,
model._weight,
model.model_type,
model.estimate_parameter,
model.calculate_err,
]
assert list(default.values()) == model_values
def test_model_prediction():
x, y, theta = create_test_data()
basis_function = Polynomial(degree=2)
train_percentage = 90
split_data = int(len(x) * (train_percentage / 100))
X_train = x[0:split_data, 0]
X_test = x[split_data::, 0]
y1 = y[0:split_data, 0]
y_test = y[split_data::, 0]
y_train = y1.copy()
y_train = np.reshape(y_train, (len(y_train), 1))
X_train = np.reshape(X_train, (len(X_train), 1))
y_test = np.reshape(y_test, (len(y_test), 1))
X_test = np.reshape(X_test, (len(X_test), 1))
model = MetaMSS(
ylag=[1, 2],
xlag=2,
maxiter=30,
n_agents=20,
basis_function=basis_function,
random_state=42,
)
model.fit(X_train=X_train, y_train=y_train, X_test=X_test, y_test=y_test)
assert_raises(Exception, model.predict, X_test=X_test, y_test=y_test[:1])
def test_default_values():
default = {
"ylag": 2,
"xlag": 2,
"model_type": "NARMAX",
"maxiter": 30,
"alpha": 23,
"g_zero": 100,
"k_agents_percent": 2,
"norm": -2,
"power": 2,
"n_agents": 10,
"p_zeros": 0.5,
"p_ones": 0.5,
"p_value": 0.05,
"estimator": "least_squares",
"extended_least_squares": False,
"lam": 0.98,
"delta": 0.01,
"offset_covariance": 0.2,
"mu": 0.01,
"eps": np.finfo(np.float64).eps,
"gama": 0.2,
"weight": 0.02,
"steps_ahead": None,
"estimate_parameter": True,
"loss_func": "metamss_loss",
"random_state": None,
}
model = MetaMSS(basis_function=Polynomial(degree=2))
model_values = [
model.ylag,
model.xlag,
model.model_type,
model.maxiter,
model.alpha,
model.g_zero,
model.k_agents_percent,
model._norm,
model._power,
model.n_agents,
model.p_zeros,
model.p_ones,
model.p_value,
model.estimator,
model._extended_least_squares,
model._lam,
model._delta,
model._offset_covariance,
model._mu,
model._eps,
model._gama,
model._weight,
model.steps_ahead,
model.estimate_parameter,
model.loss_func,
model.random_state,
]
assert list(default.values()) == model_values
def test_fit_without_information_criteria():
x, y, theta = create_test_data()
basis_function = Polynomial(degree=2)
model = FROLS(
n_terms=15, extended_least_squares=False, basis_function=basis_function
)
model.fit(X=x, y=y)
assert "info_values" not in dir(model)
def test_estimate_parameter():
x_train, x_valid, y_train, y_valid = get_siso_data(n=1000,
colored_noise=False,
sigma=0.001,
train_percentage=90)
assert_raises(
TypeError,
SimulateNARMAX,
estimate_parameter="False",
x_train=x_train,
y_train=y_train,
basis_function=Polynomial(),
)
def test_predict():
X_train, X_test, y_train, y_test = get_siso_data(
n=1000, colored_noise=False, sigma=0.0001, train_percentage=90
)
basis_function = Polynomial(degree=2)
model = MetaMSS(
ylag=[1, 2],
xlag=2,
maxiter=30,
n_agents=10,
basis_function=basis_function,
random_state=42,
)
model.fit(X_train=X_train, y_train=y_train, X_test=X_test, y_test=y_test)
yhat = model.predict(X_test=X_test, y_test=y_test)
assert_almost_equal(yhat, y_test, decimal=2)
def test_information_criteria_bic():
x, y, theta = create_test_data()
basis_function = Polynomial(degree=2)
model = FROLS(
n_terms=5,
extended_least_squares=False,
order_selection=True,
info_criteria="bic",
n_info_values=5,
ylag=[1, 2],
xlag=2,
estimator="least_squares",
basis_function=basis_function,
)
model.fit(X=x, y=y)
info_values = np.array([-1764.885, -2320.101, -2976.391, -4461.908, -72845.768])
assert_almost_equal(model.info_values[:4], info_values[:4], decimal=3)
def test_information_criteria_lilc():
x, y, theta = create_test_data()
basis_function = Polynomial(degree=2)
model = FROLS(
n_terms=5,
extended_least_squares=False,
order_selection=True,
info_criteria="lilc",
n_info_values=5,
ylag=[1, 2],
xlag=2,
estimator="least_squares",
basis_function=basis_function,
)
model.fit(X=x, y=y)
info_values = np.array([-1767.926, -2326.183, -2985.514, -4474.072, -72860.973])
assert_almost_equal(model.info_values[:4], info_values[:4], decimal=3)
def test_information_criteria_fpe():
x, y, theta = create_test_data()
basis_function = Polynomial(degree=2)
model = FROLS(
n_terms=5,
extended_least_squares=False,
order_selection=True,
info_criteria="fpe",
n_info_values=5,
ylag=[1, 2],
xlag=2,
estimator="least_squares",
basis_function=basis_function,
)
model.fit(X=x, y=y)
info_values = np.array(
[-1769.7907932, -2329.9129013, -2991.1078281, -4481.5306067, -72870.296884]
)
assert_almost_equal(model.info_values[:4], info_values[:4], decimal=3)
def test_metamss():
piv = np.array([4, 2, 7, 11, 5])
model_code = np.array(
[[1001, 0], [2002, 0], [2001, 1001]] # y(k-1) # x1(k-2) # x1(k-1)y(k-1)
)
basis_function = Polynomial(degree=2)
X_train, X_test, y_train, y_test = get_siso_data(
n=1000, colored_noise=False, sigma=0.0001, train_percentage=90
)
model = MetaMSS(
ylag=[1, 2],
xlag=2,
maxiter=30,
n_agents=20,
basis_function=basis_function,
random_state=42,
)
model.fit(X_train=X_train, y_train=y_train, X_test=X_test, y_test=y_test)
assert_array_equal(model.final_model, model_code)
def test_error_reduction_ratio():
piv = np.array([4, 2, 7, 11, 5])
model_code = np.array(
[[2002, 0], [1002, 0], [2001, 1001], [2002, 1002], [1001, 1001]]
)
basis_function = Polynomial(degree=2)
x, y, theta = create_test_data()
model = FROLS(
n_terms=5,
order_selection=True,
n_info_values=5,
info_criteria="aic",
extended_least_squares=False,
ylag=[1, 2],
xlag=2,
estimator="least_squares",
basis_function=basis_function,
)
model.fit(X=x, y=y)
assert_array_equal(model.final_model, model_code)
def test_simulate():
x_train, x_valid, y_train, y_valid = get_siso_data(n=1000,
colored_noise=False,
sigma=0.001,
train_percentage=90)
s = SimulateNARMAX(basis_function=Polynomial(), estimate_parameter=False)
# the model must be a numpy array
model = np.array([
[1001, 0], # y(k-1)
[2001, 1001], # x1(k-1)y(k-1)
[2002, 0], # x1(k-2)
])
# theta must be a numpy array of shape (n, 1) where n is the number of regressors
theta = np.array([[0.2, 0.9, 0.1]]).T
yhat = s.simulate(X_test=x_valid,
y_test=y_valid,
model_code=model,
theta=theta)
assert yhat.shape == (100, 1)
def test_default_values():
default = {
"ylag": 2,
"xlag": 2,
"order_selection": False,
"info_criteria": "aic",
"n_terms": None,
"n_info_values": 10,
"estimator": "recursive_least_squares",
"extended_least_squares": False,
"lam": 0.98,
"delta": 0.01,
"offset_covariance": 0.2,
"mu": 0.01,
"eps": np.finfo(np.float64).eps,
"gama": 0.2,
"weight": 0.02,
"model_type": "NARMAX",
}
model = FROLS(basis_function=Polynomial(degree=2))
model_values = [
model.ylag,
model.xlag,
model._order_selection,
model.info_criteria,
model.n_terms,
model.n_info_values,
model.estimator,
model._extended_least_squares,
model._lam,
model._delta,
model._offset_covariance,
model._mu,
model._eps,
model._gama,
model._weight,
model.model_type,
]
assert list(default.values()) == model_values
def test_simulate_theta():
x_train, x_valid, y_train, y_valid = get_siso_data(n=1000,
colored_noise=False,
sigma=0.001,
train_percentage=90)
s = SimulateNARMAX(basis_function=Polynomial(), estimate_parameter=True)
# the model must be a numpy array
model = np.array([
[1001, 0], # y(k-1)
[2001, 1001], # x1(k-1)y(k-1)
[2002, 0], # x1(k-2)
])
yhat = s.simulate(
X_train=x_train,
y_train=y_train,
X_test=x_valid,
y_test=y_valid,
model_code=model,
)
theta = np.array([[0.2, 0.9, 0.1]]).T
assert_almost_equal(s.theta, theta, decimal=1)
def test_validate_ylag():
assert_raises(ValueError, MetaMSS, ylag=-1, basis_function=Polynomial(degree=2))
assert_raises(ValueError, MetaMSS, ylag=1.3, basis_function=Polynomial(degree=2))
def test_validate_xlag():
assert_raises(ValueError, FROLS, xlag=-1, basis_function=Polynomial(degree=2))
assert_raises(ValueError, FROLS, xlag=1.3, basis_function=Polynomial(degree=2))
def test_info_criteria():
assert_raises(
ValueError, FROLS, info_criteria="AIC", basis_function=Polynomial(degree=2)
)
def test_calculate_error():
assert_raises(TypeError,
SimulateNARMAX,
calculate_err=1,
basis_function=Polynomial(degree=2))
