def generate_main_results():
"""Generate the main results of the experiment."""
wide_optimal = sort_tbl(
calculate_wide_optimal(results),
ovrs_order=OVRS_NAMES, clfs_order=CLFS_NAMES
)\
.set_index(['Dataset', 'Classifier', 'Metric'])\
.apply(lambda row: make_bold(row, num_decimals=3), axis=1)\
.reset_index()
wide_optimal['Dataset'] = wide_optimal['Dataset'].apply(
lambda x: x.title()
if len(x.split(' ')) == 1
else ''.join([w[0] for w in x.split(' ')])
)
mean_sem_scores = sort_tbl(
generate_mean_std_tbl_bold(*calculate_mean_sem_scores(results), maximum=True, decimals=3),
ovrs_order=OVRS_NAMES, clfs_order=CLFS_NAMES
)
mean_sem_perc_diff_scores = sort_tbl(
generate_mean_std_tbl(*calculate_mean_sem_perc_diff_scores(results, ['SMOTE', 'K-SMOTE'])),
ovrs_order=OVRS_NAMES, clfs_order=CLFS_NAMES
)
mean_sem_ranking = sort_tbl(
generate_mean_std_tbl_bold(*calculate_mean_sem_ranking(results), maximum=False),
ovrs_order=OVRS_NAMES, clfs_order=CLFS_NAMES
)
main_results_names = ('wide_optimal', 'mean_sem_scores', 'mean_sem_perc_diff_scores', 'mean_sem_ranking')
return zip(main_results_names, (wide_optimal, mean_sem_scores, mean_sem_perc_diff_scores, mean_sem_ranking))
def generate_main_results():
"""Generate the main results of the experiment."""
main_results = {}
for ratio in UNDERSAMPLING_RATIOS:
# Generate results
results = generate_results(ratio)
# Calculate results
mean_sem_scores = sort_tbl(
generate_mean_std_tbl(*calculate_mean_sem_scores(results)),
ovrs_order=OVERSAMPLERS_NAMES,
clfs_order=CLASSIFIERS_NAMES)
mean_sem_perc_diff_scores = sort_tbl(
generate_mean_std_tbl(*calculate_mean_sem_perc_diff_scores(
results, ['NO OVERSAMPLING', 'G-SMOTE'])),
ovrs_order=OVERSAMPLERS_NAMES,
clfs_order=CLASSIFIERS_NAMES)
mean_sem_ranking = sort_tbl(
generate_mean_std_tbl(*calculate_mean_sem_ranking(results)),
ovrs_order=OVERSAMPLERS_NAMES,
clfs_order=CLASSIFIERS_NAMES)
# Populate main results
main_results_names = ('mean_sem_scores', 'mean_sem_perc_diff_scores',
'mean_sem_ranking')
main_results[ratio] = zip(
main_results_names,
(mean_sem_scores, mean_sem_perc_diff_scores, mean_sem_ranking))
return main_results
def test_mean_sem_scores():
"""Test the mean scores results of experiment."""
mean_scores, sem_scores = calculate_mean_sem_scores(EXPERIMENT.results_)
assert set(mean_scores.Classifier.unique()) == set(
EXPERIMENT.classifiers_names_)
assert set(sem_scores.Classifier.unique()) == set(
EXPERIMENT.classifiers_names_)
assert set(mean_scores.columns[2:]) == set(EXPERIMENT.oversamplers_names_)
assert set(sem_scores.columns[2:]) == set(EXPERIMENT.oversamplers_names_)
assert len(mean_scores) == len(CLASSIFIERS)
assert len(sem_scores) == len(CLASSIFIERS)
def generate_main_results():
"""Generate the main results of the experiment."""
# Generate results
results = generate_results()
# Calculate results
mean_sem_scores = sort_tbl(
generate_mean_std_tbl(*calculate_mean_sem_scores(results)),
ovrs_order=OVERSAMPLERS_NAMES,
clfs_order=CLASSIFIERS_NAMES)
keys = mean_sem_scores[['Classifier', 'Metric']]
mean_sem_perc_diff_scores = []
for oversampler in ('SMOTE', 'K-MEANS SMOTE', 'SOMO', 'G-SMOTE'):
perc_diff_scores = sort_tbl(
generate_mean_std_tbl(*calculate_mean_sem_perc_diff_scores(
results, [oversampler, 'G-SOMO'])),
ovrs_order=OVERSAMPLERS_NAMES,
clfs_order=CLASSIFIERS_NAMES)
perc_diff_scores = perc_diff_scores.rename(columns={
'Difference': oversampler
}).drop(columns=['Classifier', 'Metric'])
mean_sem_perc_diff_scores.append(perc_diff_scores)
mean_sem_perc_diff_scores = pd.concat(
[keys, pd.concat(mean_sem_perc_diff_scores, axis=1)], axis=1)
mean_sem_ranking = sort_tbl(
generate_mean_std_tbl(*calculate_mean_sem_ranking(results)),
ovrs_order=OVERSAMPLERS_NAMES,
clfs_order=CLASSIFIERS_NAMES)
# Generate main results
main_results_names = ('mean_sem_scores', 'mean_sem_perc_diff_scores',
'mean_sem_ranking')
main_results = zip(
main_results_names,
(mean_sem_scores, mean_sem_perc_diff_scores, mean_sem_ranking))
return main_results