def _objective_AW2S_MDL(trial, train, changepoints, tolerance_delay):
window_size = trial.suggest_int('window_size', 10, 500)
sigma_given = trial.suggest_uniform('sigma_given', 0.1, 2)
nml_gaussian = partial(sdmdl_nml.nml_gaussian,
mu_max=1e8,
div_min=1e-8,
div_max=1e8)
complexity_gaussian = partial(sdmdl_nml.complexity_gaussian,
mu_max=1e8,
div_min=1e-8,
div_max=1e8)
retrospective_first = sdmdl.Retrospective(
h=window_size,
encoding_func=nml_gaussian,
complexity_func=complexity_gaussian,
order=0)
lnml_gaussian = partial(hsdmdl2_nml.lnml_gaussian, sigma_given=sigma_given)
retrospective_second = hsdmdl2.Retrospective(encoding_func=lnml_gaussian,
d=2,
M=5,
min_datapoints=5,
delta_0=0.05,
delta_1=0.05,
delta_2=0.05,
how_to_drop='all',
order=0,
reliability=True)
retrospective = aw2s_mdl.Retrospective(retrospective_first,
retrospective_second)
alarms = retrospective.make_alarms(train)
scores = np.zeros(len(train))
scores[alarms] = 1
F1_score, _, _ = calc_F1_score(scores,
changepoints,
tolerance_delay,
tuned_threshold=0.5)
return -F1_score
def _objective_Hierarchical(trial, train, changepoints, tolerance_delay,
order):
nml_gaussian = partial(hsdmdl2_nml.nml_gaussian)
min_datapoints = 5
# delta_0だけはいかなる場合でもチューニングしなければならない
delta_0 = trial.suggest_uniform('delta_0', 0.000001, 10.00)
if order == 1:
delta_1 = trial.suggest_uniform('delta_1', 0.000001, 1.00)
else:
delta_1 = 0.05
if order == 2:
delta_2 = trial.suggest_uniform('delta_2', 0.000001, 1.00)
else:
delta_2 = 0.05
retrospective = hsdmdl2.Retrospective(encoding_func=nml_gaussian,
d=2,
M=5,
min_datapoints=min_datapoints,
delta_0=delta_0,
delta_1=delta_1,
delta_2=delta_2,
how_to_drop='all',
order=order,
reliability=True)
alarms = retrospective.make_alarms(train)
scores = np.zeros(len(train))
scores[alarms] = 1
F1_score, _, _ = calc_F1_score(scores,
changepoints,
tolerance_delay,
tuned_threshold=0.5)
return -F1_score
def conduct_AW2S_MDL(n_trials, n_samples, dataset, changepoints,
tolerance_delay): # AW2S-MDL
# hyperparameter tuning
objective_AW2S_MDL = partial(_objective_AW2S_MDL,
train=dataset[0],
changepoints=changepoints,
tolerance_delay=tolerance_delay)
study = optuna.create_study()
study.optimize(objective_AW2S_MDL, n_trials=n_trials, n_jobs=-1)
opt_window_size = study.best_params['window_size']
opt_sigma_given = study.best_params['sigma_given']
nml_gaussian = partial(sdmdl_nml.nml_gaussian,
mu_max=1e8,
div_min=1e-8,
div_max=1e8)
complexity_gaussian = partial(sdmdl_nml.complexity_gaussian,
mu_max=1e8,
div_min=1e-8,
div_max=1e8)
retrospective_first = sdmdl.Retrospective(
h=opt_window_size,
encoding_func=nml_gaussian,
complexity_func=complexity_gaussian,
order=0)
lnml_gaussian = partial(hsdmdl2_nml.lnml_gaussian,
sigma_given=opt_sigma_given)
retrospective_second = hsdmdl2.Retrospective(encoding_func=lnml_gaussian,
d=2,
M=5,
min_datapoints=5,
delta_0=0.05,
delta_1=0.05,
delta_2=0.05,
how_to_drop='all',
order=0,
reliability=True)
retrospective = aw2s_mdl.Retrospective(retrospective_first,
retrospective_second)
# calculate metrics
calc_metrics = partial(_calc_AW2S_MDL_metrics,
dataset=dataset,
changepoints=changepoints,
tolerance_delay=tolerance_delay,
threshold=0.5,
retrospective=retrospective)
p = Pool(multi.cpu_count() - 1)
args = list(range(1, n_samples))
res = np.array(p.map(calc_metrics, args))
p.close()
# result
print("F1 score: ", np.mean(res[:, 0]), "±", np.std(res[:, 0]))
print("precision: ", np.mean(res[:, 1]), "±", np.std(res[:, 1]))
print("recall: ", np.mean(res[:, 2]), "±", np.std(res[:, 2]))
row = pd.DataFrame({
"method": ["AW2S_MDL"],
"F1_score_mean": np.mean(res[:, 0]),
"F1_score_std": np.std(res[:, 0]),
"precision_mean": np.mean(res[:, 1]),
"precision_std": np.std(res[:, 1]),
"recall_mean": np.mean(res[:, 2]),
"recall_std": np.std(res[:, 2])
})
return row
def conduct_Hierarchical(n_trials, n_samples, dataset, changepoints,
tolerance_delay, order): # Hierarchical
# hyperparameter tuning
objective_Hierarchical = partial(_objective_Hierarchical,
train=dataset[0],
changepoints=changepoints,
tolerance_delay=tolerance_delay,
order=order)
study = optuna.create_study()
study.optimize(objective_Hierarchical, n_trials=n_trials, n_jobs=-1)
opt_delta_0 = study.best_params['delta_0']
if order == 1:
opt_delta_1 = study.best_params['delta_1']
else:
opt_delta_1 = 0.05
if order == 2:
opt_delta_2 = study.best_params['delta_2']
else:
opt_delta_2 = 0.05
min_datapoints = 5
nml_gaussian = partial(hsdmdl2_nml.nml_gaussian)
retrospective = hsdmdl2.Retrospective(encoding_func=nml_gaussian,
d=2,
M=5,
min_datapoints=min_datapoints,
delta_0=opt_delta_0,
delta_1=opt_delta_1,
delta_2=opt_delta_2,
how_to_drop='all',
order=True,
reliability=True)
# calculate metrics
calc_metrics = partial(_calc_Hierarchical_metrics,
dataset=dataset,
changepoints=changepoints,
tolerance_delay=tolerance_delay,
threshold=0.5,
retrospective=retrospective,
order=order)
p = Pool(multi.cpu_count() - 1)
args = list(range(1, n_samples))
res = np.array(p.map(calc_metrics, args))
p.close()
# result
print("F1 score: ", np.mean(res[:, 0]), "±", np.std(res[:, 0]))
print("precision: ", np.mean(res[:, 1]), "±", np.std(res[:, 1]))
print("recall: ", np.mean(res[:, 2]), "±", np.std(res[:, 2]))
method_name = "Hierarchical"
if order == 0:
method_name += "_0th"
elif order == 1:
method_name += "_1st"
else:
method_name += "_2nd"
row = pd.DataFrame({
"method": [method_name],
"F1_score_mean": np.mean(res[:, 0]),
"F1_score_std": np.std(res[:, 0]),
"precision_mean": np.mean(res[:, 1]),
"precision_std": np.std(res[:, 1]),
"recall_mean": np.mean(res[:, 2]),
"recall_std": np.std(res[:, 2])
})
return row
def test_main():
np.random.seed(0)
tolerance_delay = 100
transition_period = 200
data, changepoints = one_mean_changing(transition_period=transition_period)
name = "ChangeFinder"
retrospective = changefinder.Retrospective(r=0.1, order=5, smooth=5)
_calc_metrics_draw_figure(
name, data, changepoints, tolerance_delay=tolerance_delay, retrospective=retrospective)
name = "BOCPD"
h = partial(bocpd.constant_hazard, 1000)
lik = bocpd.StudentT(0.5, 5, 5, 0)
retrospective = bocpd.Retrospective(hazard_func=h, likelihood_func=lik)
_calc_metrics_draw_figure(
name, data, changepoints, tolerance_delay=tolerance_delay, retrospective=retrospective)
name = "SDMDL_0th"
nml_gaussian = partial(sdmdl_nml.nml_gaussian, mu_max=1e8,
div_min=1e-8, div_max=1e8)
complexity_gaussian = partial(sdmdl_nml.complexity_gaussian, mu_max=1e8,
div_min=1e-8, div_max=1e8)
retrospective = sdmdl.Retrospective(h=100, encoding_func=nml_gaussian,
complexity_func=complexity_gaussian, order=0)
_calc_metrics_draw_figure(
name, data, changepoints, tolerance_delay=tolerance_delay, retrospective=retrospective)
name = "SDMDL_1st"
nml_gaussian = partial(sdmdl_nml.nml_gaussian, mu_max=1e8,
div_min=1e-8, div_max=1e8)
complexity_gaussian = partial(sdmdl_nml.complexity_gaussian, mu_max=1e8,
div_min=1e-8, div_max=1e8)
retrospective = sdmdl.Retrospective(h=100, encoding_func=nml_gaussian,
complexity_func=complexity_gaussian, order=1)
_calc_metrics_draw_figure(
name, data, changepoints, tolerance_delay=tolerance_delay, retrospective=retrospective)
name = "SDMDL_2nd"
nml_gaussian = partial(sdmdl_nml.nml_gaussian, mu_max=1e8,
div_min=1e-8, div_max=1e8)
complexity_gaussian = partial(sdmdl_nml.complexity_gaussian, mu_max=1e8,
div_min=1e-8, div_max=1e8)
retrospective = sdmdl.Retrospective(h=100, encoding_func=nml_gaussian,
complexity_func=complexity_gaussian, order=2)
_calc_metrics_draw_figure(
name, data, changepoints, tolerance_delay=tolerance_delay, retrospective=retrospective)
name = "Hierarchical_SCAW1_0th"
lnml_gaussian = partial(hsdmdl1_nml.lnml_gaussian)
retrospective = hsdmdl1.Retrospective(encoding_func=lnml_gaussian, d=2, min_datapoints=5, delta_0=0.05,
delta_1=0.05, delta_2=0.05, how_to_drop='all', order=0, reliability=True)
_calc_metrics_draw_figure(
name, data, changepoints, tolerance_delay=tolerance_delay, retrospective=retrospective)
name = "Hierarchical_SCAW1_1st"
lnml_gaussian = partial(hsdmdl1_nml.lnml_gaussian)
retrospective = hsdmdl1.Retrospective(encoding_func=lnml_gaussian, d=2, min_datapoints=5, delta_0=0.05,
delta_1=0.05, delta_2=0.05, how_to_drop='all', order=1, reliability=True)
_calc_metrics_draw_figure(
name, data, changepoints, tolerance_delay=tolerance_delay, retrospective=retrospective)
name = "Hierarchical_SCAW1_2nd"
lnml_gaussian = partial(hsdmdl1_nml.lnml_gaussian)
retrospective = hsdmdl1.Retrospective(encoding_func=lnml_gaussian, d=2, min_datapoints=5, delta_0=0.05,
delta_1=0.05, delta_2=0.05, how_to_drop='all', order=2, reliability=True)
_calc_metrics_draw_figure(
name, data, changepoints, tolerance_delay=tolerance_delay, retrospective=retrospective)
name = "Hierarchical_SCAW2_0th"
nml_gaussian = partial(hsdmdl2_nml.nml_gaussian)
retrospective = hsdmdl2.Retrospective(encoding_func=nml_gaussian, d=2, M=5, min_datapoints=5, delta_0=0.05,
delta_1=0.05, delta_2=0.05, how_to_drop='all', order=0, reliability=True)
_calc_metrics_draw_figure(
name, data, changepoints, tolerance_delay=tolerance_delay, retrospective=retrospective)
name = "Hierarchical_SCAW2_1st"
nml_gaussian = partial(hsdmdl2_nml.nml_gaussian)
retrospective = hsdmdl2.Retrospective(encoding_func=nml_gaussian, d=2, M=5, min_datapoints=5, delta_0=0.05,
delta_1=0.05, delta_2=0.05, how_to_drop='all', order=1, reliability=True)
_calc_metrics_draw_figure(
name, data, changepoints, tolerance_delay=tolerance_delay, retrospective=retrospective)
name = "Hierarchical_SCAW2_2nd"
nml_gaussian = partial(hsdmdl2_nml.nml_gaussian)
retrospective = hsdmdl2.Retrospective(encoding_func=nml_gaussian, d=2, M=5, min_datapoints=5, delta_0=0.05,
delta_1=0.05, delta_2=0.05, how_to_drop='all', order=2, reliability=True)
_calc_metrics_draw_figure(
name, data, changepoints, tolerance_delay=tolerance_delay, retrospective=retrospective)
name = "FW2S_MDL"
nml_gaussian = partial(sdmdl_nml.nml_gaussian, mu_max=1e8,
div_min=1e-8, div_max=1e8)
complexity_gaussian = partial(sdmdl_nml.complexity_gaussian, mu_max=1e8,
div_min=1e-8, div_max=1e8)
retrospective_first = sdmdl.Retrospective(h=100, encoding_func=nml_gaussian,
complexity_func=complexity_gaussian, order=0)
nml_gaussian = partial(sdmdl_nml.nml_gaussian, mu_max=1e8,
div_min=1e-8, div_max=1e8)
complexity_gaussian = partial(sdmdl_nml.complexity_gaussian, mu_max=1e8,
div_min=1e-8, div_max=1e8)
retrospective_second = sdmdl.Retrospective(h=100, encoding_func=nml_gaussian,
complexity_func=complexity_gaussian, order=0)
retrospective = fw2s_mdl.Retrospective(
retrospective_first, retrospective_second)
_calc_metrics_draw_figure(
name, data, changepoints, tolerance_delay=tolerance_delay, retrospective=retrospective)
name = "AW2S_MDL"
nml_gaussian = partial(sdmdl_nml.nml_gaussian, mu_max=1e8,
div_min=1e-8, div_max=1e8)
complexity_gaussian = partial(sdmdl_nml.complexity_gaussian, mu_max=1e8,
div_min=1e-8, div_max=1e8)
retrospective_first = sdmdl.Retrospective(h=100, encoding_func=nml_gaussian,
complexity_func=complexity_gaussian, order=0)
lnml_gaussian = partial(hsdmdl2_nml.lnml_gaussian, sigma_given=0.3)
retrospective_second = hsdmdl2.Retrospective(encoding_func=lnml_gaussian, d=2, M=5, min_datapoints=5, delta_0=0.05,
delta_1=0.05, delta_2=0.05, how_to_drop='all', order=0, reliability=True)
retrospective = aw2s_mdl.Retrospective(
retrospective_first, retrospective_second)
_calc_metrics_draw_figure(
name, data, changepoints, tolerance_delay=tolerance_delay, retrospective=retrospective)