def test_mfdcca_save_load():
mfdcca = fathon.MFDCCA(fu.toAggregated(ts), fu.toAggregated(ts))
# save and load with empty results
mfdcca.saveObject(get_object_path('mfdcca_obj'))
n_load = fu.getObjectMember(get_object_path('mfdcca_obj', ext=True), 'n')
F_load = fu.getObjectMember(get_object_path('mfdcca_obj', ext=True), 'F')
q_list_load = fu.getObjectMember(get_object_path('mfdcca_obj', ext=True),
'qList')
list_h_load = fu.getObjectMember(get_object_path('mfdcca_obj', ext=True),
'listH')
assert np.array_equal(n_load, [])
assert np.array_equal(F_load, [])
assert np.array_equal(q_list_load, [])
assert np.array_equal(list_h_load, [])
#save and load with results
n, F = mfdcca.computeFlucVec(fu.linRangeByStep(10, 500),
fu.linRangeByStep(-1, 1))
H, I = mfdcca.fitFlucVec(100, 300)
mfdcca.saveObject(get_object_path('mfdcca_obj'))
n_load = fu.getObjectMember(get_object_path('mfdcca_obj', ext=True), 'n')
F_load = fu.getObjectMember(get_object_path('mfdcca_obj', ext=True), 'F')
q_list_load = fu.getObjectMember(get_object_path('mfdcca_obj', ext=True),
'qList')
list_h_load = fu.getObjectMember(get_object_path('mfdcca_obj', ext=True),
'listH')
assert np.array_equal(n_load, n)
assert np.array_equal(F_load, F)
assert np.array_equal(q_list_load, fu.linRangeByStep(-1, 1))
assert np.array_equal(list_h_load, H)
# MFDFA from file
mfdcca_2 = fathon.MFDCCA(get_object_path('mfdcca_obj', ext=True))
H_2, I_2 = mfdcca_2.fitFlucVec(100, 300)
assert np.array_equal(H_2, H)
assert np.array_equal(I_2, I)
def test_mat_dfa_mf():
mf_dfa = fathon.DFA(fu.toAggregated(mf))
n_mf, F_mf = mf_dfa.computeFlucVec(scales, revSeg=False, polOrd=1)
idxs = get_idxs(n_mf, scales)
n_mf = n_mf[idxs]
F_mf = F_mf[idxs]
H, _ = np.polyfit(np.log2(n_mf), np.log2(F_mf), 1)
assert math.isclose(H, 0.77, rel_tol=1e-2, abs_tol=0)
def test_mat_dfa_wn():
w_dfa = fathon.DFA(fu.toAggregated(wn))
n_w, F_w = w_dfa.computeFlucVec(scales, revSeg=False, polOrd=1)
idxs = get_idxs(n_w, scales)
n_w = n_w[idxs]
F_w = F_w[idxs]
H, _ = np.polyfit(np.log2(n_w), np.log2(F_w), 1)
assert math.isclose(H, 0.45, rel_tol=1e-2, abs_tol=0)
def test_ht_save_load():
ht = fathon.HT(fu.toAggregated(ts))
# save and load with empty results
ht.saveObject(get_object_path('ht_obj'))
ht_load = fu.getObjectMember(get_object_path('ht_obj', ext=True), 'ht')
assert np.array_equal(ht_load, [])
#save and load with results
ht_mtx = ht.computeHt([100, 300])
ht.saveObject(get_object_path('ht_obj'))
ht_load = fu.getObjectMember(get_object_path('ht_obj', ext=True), 'ht')
assert np.array_equal(ht_load, ht_mtx)
def test_mat_mfdfa_wn():
w_mfdfa = fathon.MFDFA(fu.toAggregated(wn))
n_w, F_w = w_mfdfa.computeFlucVec(scales,
qList=q_list,
revSeg=False,
polOrd=1)
idxs = get_idxs(n_w, scales)
n_w = n_w[idxs]
F_w_vec = np.zeros((len(q_list), len(idxs)))
for i in range(len(q_list)):
F_w_vec[i] = F_w[i, idxs]
Hq = []
for i in range(len(q_list)):
Hq.append(np.polyfit(np.log2(n_w), np.log2(F_w_vec[i]), 1)[0])
np.testing.assert_allclose(
Hq, [0.4583, 0.4555, 0.4546, 0.4515, 0.4445, 0.4340],
rtol=1e-4,
atol=0)
def test_mat_mfdfa_mf():
mf_mfdfa = fathon.MFDFA(fu.toAggregated(mf))
n_mf, F_mf = mf_mfdfa.computeFlucVec(scales,
qList=q_list,
revSeg=False,
polOrd=1)
idxs = get_idxs(n_mf, scales)
n_mf = n_mf[idxs]
F_mf_vec = np.zeros((len(q_list), len(idxs)))
for i in range(len(q_list)):
F_mf_vec[i] = F_mf[i, idxs]
Hq = []
for i in range(len(q_list)):
Hq.append(np.polyfit(np.log2(n_mf), np.log2(F_mf_vec[i]), 1)[0])
np.testing.assert_allclose(
Hq, [1.4477, 1.3064, 1.0823, 0.8846, 0.6606, 0.5174],
rtol=1e-4,
atol=0)
def test_mat_mfdfa_mn():
mn_mfdfa = fathon.MFDFA(fu.toAggregated(mn))
n_mn, F_mn = mn_mfdfa.computeFlucVec(scales,
qList=q_list,
revSeg=False,
polOrd=1)
idxs = get_idxs(n_mn, scales)
n_mn = n_mn[idxs]
F_mn_vec = np.zeros((len(q_list), len(idxs)))
for i in range(len(q_list)):
F_mn_vec[i] = F_mn[i, idxs]
Hq = []
for i in range(len(q_list)):
Hq.append(np.polyfit(np.log2(n_mn), np.log2(F_mn_vec[i]), 1)[0])
np.testing.assert_allclose(
Hq, [0.7542, 0.7392, 0.7301, 0.7240, 0.7149, 0.7023],
rtol=1e-4,
atol=0)
def test_dfa_save_load():
dfa = fathon.DFA(fu.toAggregated(ts))
# save and load with empty results
dfa.saveObject(get_object_path('dfa_obj'))
n_load = fu.getObjectMember(get_object_path('dfa_obj', ext=True), 'n')
F_load = fu.getObjectMember(get_object_path('dfa_obj', ext=True), 'F')
assert np.array_equal(n_load, [])
assert np.array_equal(F_load, [])
#save and load with results
n, F = dfa.computeFlucVec(fu.linRangeByStep(10, 500))
H, I = dfa.fitFlucVec(100, 300)
dfa.saveObject(get_object_path('dfa_obj'))
n_load = fu.getObjectMember(get_object_path('dfa_obj', ext=True), 'n')
F_load = fu.getObjectMember(get_object_path('dfa_obj', ext=True), 'F')
assert np.array_equal(n_load, n)
assert np.array_equal(F_load, F)
# DFA from file
dfa_2 = fathon.DFA(get_object_path('dfa_obj', ext=True))
H_2, I_2 = dfa_2.fitFlucVec(100, 300)
assert H_2 == H
assert I_2 == I
# REGRESSION TESTS # ---------------- # - the element of the array tested in tests 2-5,7,8 has no # particular meaning, it was just chosen randomly. # - the numbers to which results are tested against have # been produced by fathon and have many significant figures # due to the fact that the precision of the testing machine # is not known. ##### # Time series for regression tests ##### # co2 residuals after yearly cycle removal ts1 = np.loadtxt(os.path.join(TESTS_PATH, 'ts1.txt')) ts1 = fu.toAggregated(ts1) # other co2 residuals after yearly cycle removal ts2 = np.loadtxt(os.path.join(TESTS_PATH, 'ts2.txt')) ts2 = fu.toAggregated(ts2) # multifractal data ts3 = np.loadtxt(os.path.join(TESTS_PATH, 'ts3.txt')) ts3 = fu.toAggregated(ts3) ##### ##### # Regression test 1 # It tests if the Hurst exponent of `ts1` is correct ##### def test_dfa_lin_step():
tick_bar_df = testClass.get_concat_data(testClass._bars_dict)['tick_bars']
volume_bar_df = testClass.get_concat_data(testClass._bars_dict)['volume_bars']
usd_volume_bar_df = testClass.get_concat_data(testClass._bars_dict)['usd_volume_bars']
calendar_bar_df = testClass.get_concat_data(testClass._bars_dict)['calendar_bars']
vr = returns(volume_bar_df.micro_price_close).replace([np.inf, -np.inf], 0) # volume
tr = returns(tick_bar_df.micro_price_close).replace([np.inf, -np.inf], 0) # tick
dr = returns(usd_volume_bar_df.micro_price_close).dropna().replace([np.inf, -np.inf], 0) # usd volume
df_ret = returns(calendar_bar_df.micro_price_close).dropna().replace([np.inf, -np.inf], 0) # calendar
bar_returns[date] = {'tick': tr,
'volume': vr,
'dollar': dr,
'calendar': df_ret}
for j, i in itertools.product(['tick', 'volume', 'dollar', 'calendar'], dates):
data = (bar_returns[i][j])
a = fu.toAggregated(np.asanyarray(data))
# MFDFA Computations
pymfdfa = fathon.MFDFA(a)
n, F = pymfdfa.computeFlucVec(winSizes, qs, revSeg=revSeg, polOrd=polOrd)
mfdfa_n_F_dict[j][i] = dict(zip(n, F))
# dictionary to match all the n and F values this could be
# more efficient
# get the list values of H and intercept
list_H, list_H_intercept = pymfdfa.fitFlucVec() # same for H values
mfdfa_H_dict[j][i] = [list_H, list_H_intercept]
# get the mass exponents
tau = pymfdfa.computeMassExponents()
mfdfa_tau_dict[j][i] = tau
def test_dcca_save_load():
dcca = fathon.DCCA(fu.toAggregated(ts), fu.toAggregated(ts))
# save and load with empty results
dcca.saveObject(get_object_path('dcca_obj'))
n_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True), 'n')
F_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True), 'F')
n_rho_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'nRho')
rho_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True), 'rho')
n_thr_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'nThr')
conf_up_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'confUp')
conf_down_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'confDown')
assert np.array_equal(n_load, [])
assert np.array_equal(F_load, [])
assert np.array_equal(n_rho_load, [])
assert np.array_equal(rho_load, [])
assert np.array_equal(n_thr_load, [])
assert np.array_equal(conf_up_load, [])
assert np.array_equal(conf_down_load, [])
#save and load with results
n, F = dcca.computeFlucVec(fu.linRangeByStep(10, 200))
H, I = dcca.fitFlucVec(100, 150)
dcca.saveObject(get_object_path('dcca_obj'))
n_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True), 'n')
F_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True), 'F')
n_rho_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'nRho')
rho_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True), 'rho')
n_thr_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'nThr')
conf_up_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'confUp')
conf_down_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'confDown')
assert np.array_equal(n_load, n)
assert np.array_equal(F_load, F)
assert np.array_equal(n_rho_load, [])
assert np.array_equal(rho_load, [])
assert np.array_equal(n_thr_load, [])
assert np.array_equal(conf_up_load, [])
assert np.array_equal(conf_down_load, [])
# DCCA from file
dcca_2 = fathon.DCCA(get_object_path('dcca_obj', ext=True))
H_2, I_2 = dcca_2.fitFlucVec(100, 150)
assert np.array_equal(H_2, H)
assert np.array_equal(I_2, I)
# rho
n_rho, rho = dcca.computeRho(fu.linRangeByStep(10, 30))
dcca.saveObject(get_object_path('dcca_obj'))
n_rho_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'nRho')
rho_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True), 'rho')
n_thr_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'nThr')
conf_up_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'confUp')
conf_down_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'confDown')
assert np.array_equal(n_load, n)
assert np.array_equal(F_load, F)
assert np.array_equal(n_rho_load, n_rho)
assert np.array_equal(rho_load, rho)
assert np.array_equal(n_thr_load, [])
assert np.array_equal(conf_up_load, [])
assert np.array_equal(conf_down_load, [])
# thresholds
n_thr, conf_up, conf_down = dcca.rhoThresholds(len(ts),
fu.linRangeByStep(10, 30),
10, 0.95)
dcca.saveObject(get_object_path('dcca_obj'))
n_thr_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'nThr')
conf_up_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'confUp')
conf_down_load = fu.getObjectMember(get_object_path('dcca_obj', ext=True),
'confDown')
assert np.array_equal(n_load, n)
assert np.array_equal(F_load, F)
assert np.array_equal(n_rho_load, n_rho)
assert np.array_equal(rho_load, rho)
assert np.array_equal(n_thr_load, n_thr)
assert np.array_equal(conf_up_load, conf_up)
assert np.array_equal(conf_down_load, conf_down)
import numpy as np
import matplotlib.pyplot as plt
import fathon
from fathon import fathonUtils as fu
print('This is fathon v{}'.format(fathon.__version__))
a = np.random.randn(10000)
b = np.random.randn(10000)
a = fu.toAggregated(a)
b = fu.toAggregated(b)
pymfdcca = fathon.MFDCCA(a, b)
winSizes = fu.linRangeByStep(10, 2000)
qs = np.arange(-3, 4, 0.1)
revSeg = True
polOrd = 1
n, F = pymfdcca.computeFlucVec(winSizes, qs, revSeg=revSeg, polOrd=polOrd)
list_H, list_H_intercept = pymfdcca.fitFlucVec()
plt.plot(np.log(n), np.log(F[0, :]), 'ro')
plt.plot(np.log(n),
list_H_intercept[0] + list_H[0] * np.log(n),
'k-',
label='h_{:.1f} = {:.2f}'.format(qs[0], list_H[0]))
half_idx = int(len(qs) / 2)
plt.plot(np.log(n), np.log(F[half_idx, :]), 'co')
