def test_allocate_mcos():
"""
Test the estimates of the optimal allocation using the Monte Carlo optimization selection
"""
nco = NCO()
# Covariance matrix, mean vector and other variables for the method
np.random.seed(0)
mu_vec = np.array([0, 0.1, 0.2, 0.3])
cov_mat = np.array([[1, 0.1, 0.2, 0.3], [0.1, 1, 0.1, 0.2],
[0.2, 0.1, 1, 0.1], [0.3, 0.2, 0.1, 1]])
num_obs = 100
num_sims = 2
kde_bwidth = 0.25
min_var_portf = True
lw_shrinkage = False
# Alternative set of values
min_var_portf_alt = False
kde_bwidth_alt = 0
# Expected weights for minimum variance allocation
w_cvo_expected = pd.DataFrame(
[[0.249287, 0.256002, 0.242593, 0.252118],
[0.257547, 0.265450, 0.242453, 0.234551]])
w_nco_expected = pd.DataFrame(
[[0.248396, 0.243172, 0.250751, 0.257680],
[0.257547, 0.265450, 0.242453, 0.234551]])
# Expected weights for maximum Sharpe ratio allocation
w_cvo_sr_expected = pd.DataFrame(
[[-1.081719, 1.810936, 1.218067, 3.978880],
[-2.431651, 0.594868, -0.210175, 5.117628]])
w_nco_sr_expected = pd.DataFrame(
[[-1.060835, 1.910503, 1.315026, 3.908128],
[-0.937168, 1.886158, -0.389275, 4.884809]])
# Finding the optimal weights for minimum variance
w_cvo, w_nco = nco.allocate_mcos(mu_vec, cov_mat, num_obs, num_sims,
kde_bwidth, min_var_portf,
lw_shrinkage)
# Finding the optimal weights for maximum Sharpe ratio
w_cvo_sr, w_nco_sr = nco.allocate_mcos(mu_vec, cov_mat, num_obs,
num_sims, kde_bwidth_alt,
min_var_portf_alt, lw_shrinkage)
# Testing if the optimal allocation simulations are right
np.testing.assert_almost_equal(np.array(w_cvo),
np.array(w_cvo_expected),
decimal=4)
np.testing.assert_almost_equal(np.array(w_nco),
np.array(w_nco_expected),
decimal=4)
np.testing.assert_almost_equal(np.array(w_cvo_sr),
np.array(w_cvo_sr_expected),
decimal=4)
np.testing.assert_almost_equal(np.array(w_nco_sr),
np.array(w_nco_sr_expected),
decimal=4)
def test_allocate_mcos():
"""
Test the estimates of the optimal allocation using the Monte Carlo optimization selection
"""
nco = NCO()
# Covariance matrix, mean vector and other variables for the method
np.random.seed(0)
mu_vec = np.array([0, 0.1, 0.2, 0.3])
cov_mat = np.array([[1, 0.1, 0.2, 0.3],
[0.1, 1, 0.1, 0.2],
[0.2, 0.1, 1, 0.1],
[0.3, 0.2, 0.1, 1]])
num_obs = 100
num_sims = 2
kde_bwidth = 0.25
min_var_portf = True
lw_shrinkage = False
# Alternative set of values
min_var_portf_alt = False
kde_bwidth_alt = 0
# Expected weights for minimum variance allocation
# Second line in the below test was changed after v.0.11.3 from [0.257547, 0.265450, 0.242453, 0.234551]
# to [0.303161, 0.256327, 0.177548, 0.262963]
# due to scikit-learn changing np.random outputs
w_cvo_expected = pd.DataFrame([[0.249287, 0.256002, 0.242593, 0.252118],
[0.303161, 0.256327, 0.177548, 0.262963]])
# Second line in the below test was changed after v.0.11.3 from [0.257547, 0.265450, 0.242453, 0.234551]
# to [0.301132, 0.247042, 0.198597, 0.25323]
# due to scikit-learn changing np.random outputs
w_nco_expected = pd.DataFrame([[0.248396, 0.243172, 0.250751, 0.257680],
[0.301132, 0.247042, 0.198597, 0.25323]])
# Expected weights for maximum Sharpe ratio allocation
# Values in the below test was changed after v.0.11.3 from [[-1.081719, 1.810936, 1.218067, 3.978880]
# [-2.431651, 0.594868, -0.210175, 5.117628]]
# to [[-0.128849, -0.326671, 0.870183, 2.020053]
# [-2.095454, -0.7403, 2.390951, 1.793756]]
# due to scikit-learn changing np.random outputs
w_cvo_sr_expected = pd.DataFrame([[-0.128849, -0.326671, 0.870183, 2.020053],
[-2.095454, -0.7403, 2.390951, 1.793756]])
# Values in the below test was changed after v.0.11.3 from [[-1.060835, 1.910503, 1.315026, 3.908128]
# [-0.937168, 1.886158, -0.389275, 4.884809]]
# to [[-0.204089, -0.050088, 0.912494, 1.983382]
# [-2.036145, -0.769203, 2.677561, 1.347365]]
# due to scikit-learn changing np.random outputs
w_nco_sr_expected = pd.DataFrame([[-0.204089, -0.050088, 0.912494, 1.983382],
[-2.036145, -0.769203, 2.677561, 1.347365]])
# Finding the optimal weights for minimum variance
w_cvo, w_nco = nco.allocate_mcos(mu_vec, cov_mat, num_obs, num_sims, kde_bwidth, min_var_portf, lw_shrinkage)
# Finding the optimal weights for maximum Sharpe ratio
w_cvo_sr, w_nco_sr = nco.allocate_mcos(mu_vec, cov_mat, num_obs, num_sims, kde_bwidth_alt, min_var_portf_alt, lw_shrinkage)
# Testing if the optimal allocation simulations are right
np.testing.assert_almost_equal(np.array(w_cvo), np.array(w_cvo_expected), decimal=6)
np.testing.assert_almost_equal(np.array(w_nco), np.array(w_nco_expected), decimal=6)
np.testing.assert_almost_equal(np.array(w_cvo_sr), np.array(w_cvo_sr_expected), decimal=6)
np.testing.assert_almost_equal(np.array(w_nco_sr), np.array(w_nco_sr_expected), decimal=6)