def generate_statistical_results():
"""Generate the statistical results of the experiment."""
# Combine experiments objects
results = []
for ratio in UNDERSAMPLING_RATIOS:
# Generate results
partial_results = generate_results(ratio)
# Extract results
cols = partial_results.columns
partial_results = partial_results.reset_index()
partial_results['Dataset'] = partial_results['Dataset'].apply(lambda name: f'{name}({ratio})')
partial_results.set_index(['Dataset', 'Oversampler', 'Classifier', 'params'], inplace=True)
partial_results.columns = cols
results.append(partial_results)
# Combine results
results = combine_results(*results)
# Calculate statistical results
friedman_test = sort_tbl(generate_pvalues_tbl(apply_friedman_test(results)), ovrs_order=OVERSAMPLERS_NAMES, clfs_order=CLASSIFIERS_NAMES)
holms_test = sort_tbl(generate_pvalues_tbl(apply_holms_test(results, control_oversampler='G-SMOTE')), ovrs_order=OVERSAMPLERS_NAMES[:-1], clfs_order=CLASSIFIERS_NAMES)
statistical_results_names = ('friedman_test', 'holms_test')
statistical_results = zip(statistical_results_names, (friedman_test, holms_test))
return statistical_results
def generate_statistical_results():
"""Generate the statistical results of the experiment."""
friedman_test = sort_tbl(generate_pvalues_tbl(apply_friedman_test(results)), ovrs_order=OVRS_NAMES, clfs_order=CLFS_NAMES)
holms_test = sort_tbl(generate_pvalues_tbl_bold(apply_holms_test(results, control_oversampler='K-SMOTE')), ovrs_order=OVRS_NAMES[:-1], clfs_order=CLFS_NAMES)
statistical_results_names = ('friedman_test', 'holms_test')
statistical_results = zip(statistical_results_names, (friedman_test, holms_test))
return statistical_results
def generate_statistical_results():
"""Generate the statistical results of the experiment."""
# Generate results
results = generate_results()
# Calculate statistical results
friedman_test = sort_tbl(generate_pvalues_tbl(
apply_friedman_test(results)),
ovrs_order=OVERSAMPLERS_NAMES,
clfs_order=CLASSIFIERS_NAMES)
holms_test = sort_tbl(generate_pvalues_tbl(
apply_holms_test(results, control_oversampler='G-SOMO')),
ovrs_order=OVERSAMPLERS_NAMES[:-1],
clfs_order=CLASSIFIERS_NAMES)
# Generate statistical results
statistical_results_names = ('friedman_test', 'holms_test')
statistical_results = zip(statistical_results_names,
(friedman_test, holms_test))
return statistical_results
def test_friedman_test():
"""Test the results of friedman test."""
friedman_test = apply_friedman_test(EXPERIMENT.results_, alpha=0.05)
assert set(friedman_test.Classifier.unique()) == set(
EXPERIMENT.classifiers_names_)
assert len(friedman_test) == len(CLASSIFIERS)