feat_imp = plot_utils.plot_feature_imp(train_test_data_cent[0], train_test_data_cent[1], model_hdl)
for i_label, label in enumerate(leg_labels):
feat_imp[i_label].savefig(f'{PLOT_DIR}/train_test_out/feature_imp_training_{bin_df}_{label}.pdf')
feat_imp[3].savefig(f'{PLOT_DIR}/train_test_out/feature_imp_training_{bin_df}_all.pdf')
plot_utils.plot_roc_train_test(
train_test_data_cent[3],
test_y_score, train_test_data_cent[1],
train_y_score, labels=leg_labels, multi_class_opt="ovr")
plt.savefig(f'{PLOT_DIR}/train_test_out/roc_train_test_{bin_df}.pdf')
plt.close('all')
if COMPUTE_SCORES_FROM_EFF:
pass
# get scores corresponding to BDT prompt efficiencies using test set
eff_selected = np.arange(0.1, MAX_EFF, 0.01)
eff, score = analysis_utils.bdt_efficiency_array(
train_test_data_cent[3], test_y_score, keep_lower=False)
score_list = []
for eff_val in eff_selected:
interp = scipy.interpolate.InterpolatedUnivariateSpline(score, eff[2]-eff_val)
score_list.append(interp.roots()[0])
score_array = np.array(score_list)
score_eff_arrays_dict[bin_df] = score_array
# write test set data frame
# train_test_data_cent[2]['model_output'] = test_y_score
# train_test_data_cent[2]['y_true'] = train_test_data_cent[3]
# train_test_data_cent_tmp = train_test_data_cent[2].query(f'y_true > 1.5 and ct >= {ct_bins_df[0]} and ct < {ct_bins_df[1]}')
# train_test_data_cent_tmp.to_parquet(f'df/mc_prompt_{bin_df}.parquet.gzip', compression='gzip')
DIGITS = pd.DataFrame(DIGITS_DATA.data[:, 0:10]) # pylint: disable=E1101
Y_DIGITS = DIGITS_DATA.target # pylint: disable=E1101
SIG_DF = DIGITS[Y_DIGITS == 1]
BKG_DF = DIGITS[Y_DIGITS == 0]
TRAIN_SET, TEST_SET, Y_TRAIN, Y_TEST = train_test_split(
DIGITS, Y_DIGITS, test_size=0.5, random_state=42)
DATA = [TRAIN_SET, Y_TRAIN, TEST_SET, Y_TEST]
# --------------------------------------------
# training and testing
INPUT_MODEL = xgb.XGBClassifier()
MODEL = ModelHandler(INPUT_MODEL)
MODEL.train_test_model(DATA)
Y_PRED = MODEL.predict(DATA[2])
Y_PRED_TRAIN = MODEL.predict(DATA[0])
EFFICIENCY, THRESHOLD = analysis_utils.bdt_efficiency_array(DATA[3], Y_PRED, n_points=10)
# --------------------------------------------
def test_plot_distr():
"""
Test the feature distribution plot
"""
assert isinstance(plot_utils.plot_distr(
[SIG_DF, BKG_DF], SIG_DF.columns), np.ndarray)
def test_plot_corr():
"""
Test the correlation matrix plot
"""
if OPTIMIZE:
model_handler.optimize_params_bayes(data,
HYPERPARAMS_RANGE,
'roc_auc',
init_points=10,
n_iter=10)
model_handler.train_test_model(data)
print("train test model")
print(
f'--- model trained and tested in {((time.time() - part_time) / 60):.2f} minutes ---\n'
)
y_pred = model_handler.predict(data[2])
data[2].insert(0, 'score', y_pred)
eff, tsd = analysis_utils.bdt_efficiency_array(
data[3], y_pred, n_points=1000)
score_from_eff_array = analysis_utils.score_from_efficiency_array(
data[3], y_pred, FIX_EFF_ARRAY)
fixed_eff_array = np.vstack(
(FIX_EFF_ARRAY, score_from_eff_array))
if SIGMA_MC:
ml_analysis.MC_sigma_array(data, fixed_eff_array,
cclass, ptbin, ctbin, split)
ml_analysis.save_ML_analysis(model_handler,
fixed_eff_array,
cent_class=cclass,
pt_range=ptbin,
ct_range=ctbin,
split=split)