def test_dataset_analysis_report(self):
da = DatasetAnalysis(REGRESSION,
class_count_threshold=10,
outliers_n_sigma=(3, 5),
nan_threshold_per_col=(0.1, 0.5, 0.9),
nan_threshold_per_sample=(0.1, 0.9))
report = da.dataset_analysis_report(dataset=self.dataset)
pprint(report.to_dict(printable=True))
pprint(report.to_dict_meaning())
pprint(report.plot())
def _create_dataset_analysis_report(self,
**kwargs) -> DatasetAnalysisReport:
self.dataset_analysis = DatasetAnalysis(
problem_Type=CLASSIFICATION if self.
is_classification else REGRESSION)
return self.create_dataset_analysis_report(
train=self.train,
test=self.test,
is_classification=self.is_classification,
dataset_analysis=self.dataset_analysis)
def test_count_outliers(self):
da = DatasetAnalysis(REGRESSION,
class_count_threshold=10,
outliers_n_sigma=(3, 5),
nan_threshold_per_col=(0.1, 0.5, 0.9),
nan_threshold_per_sample=(0.1, 0.5, 0.9))
out = da.count_outliers(self.dataset)
print("count_outliers:")
pprint(out)
self.assertTrue(9 in out)
self.assertTrue('3-sigma' in out[9])
self.assertTrue('5-sigma' in out[9])
self.assertTrue(out[9]['5-sigma']['n_outliers'] == 1)
self.assertTrue(out[9]['3-sigma']['expected_outliers'] == 2)
self.assertTrue(out[9]['5-sigma']['expected_outliers'] == 0)
self.assertTrue('target' in out)
self.assertTrue('3-sigma' not in out['target'])
self.assertTrue('5-sigma' in out['target'])
self.assertTrue(out['target']['5-sigma']['n_outliers'] == 1)
def test_count_unique_classes(self):
da = DatasetAnalysis(CLASSIFICATION,
class_count_threshold=10,
outliers_n_sigma=(3, 5),
nan_threshold_per_col=(0.1, 0.5, 0.9),
nan_threshold_per_sample=(0.1, 0.5, 0.9))
out = da.count_unique_classes(self.dataset)
print("count_unique_classes:")
pprint(out)
self.assertTrue(0 in out)
self.assertTrue(10.0 in out[0])
self.assertTrue(out[0][10.0] == 1)
self.assertTrue('target' in out)
self.assertTrue(10.0 in out['target'])
self.assertTrue(out['target'][10.0] == 1)
for feature in out:
for value in out[feature].values():
self.assertLessEqual(value, da._class_count_threshold)
def create_dataset_analysis_report(
cls,
train: DMD,
is_classification,
test: DMD = None,
dataset_analysis: DatasetAnalysis = None,
**kwargs) -> DatasetAnalysisReport:
"""
Create dataset analysis report by analyzing train data
Args:
train - train data
test - test data, if available
is_classification - whether the target is categorical.
Returns:
dataset analysis report
"""
dataset_analysis = dataset_analysis or DatasetAnalysis(
problem_Type=CLASSIFICATION if is_classification else REGRESSION)
report = dataset_analysis.dataset_analysis_report(train=train,
test=test)
return report
def test_count_missing_values(self):
da = DatasetAnalysis(REGRESSION,
class_count_threshold=10,
outliers_n_sigma=(3, 5),
nan_threshold_per_col=(0.1, 0.5, 0.9),
nan_threshold_per_sample=(0.1, 0.9))
nan_cols, nan_rows = da.count_missing_values(self.dataset)
print("count_missing_values cols:")
pprint(nan_cols)
print("count_missing_values rows:")
pprint(nan_rows)
for th in da._nan_threshold_per_feature:
self.assertTrue(th in nan_cols)
for th in da._nan_threshold_per_sample:
self.assertTrue(th in nan_rows)
self.assertEqual(nan_cols, {
0.1: {
1: 1.0,
2: 0.7
},
0.5: {
1: 1.0,
2: 0.7
},
0.9: {
1: 1.0
}
})
self.assertEqual(nan_rows[0.9], {10: 1.0})
self.assertEqual(len(nan_rows[0.1]), self.dataset.n_samples)
for th, vdict in nan_rows.items():
print(th, vdict)
self.assertTrue(th <= max(vdict.values()))
def create_dataset_analysis_report(cls, train: DMD, is_classification,
**kwargs) -> DatasetAnalysisReport:
da = DatasetAnalysis(
problem_Type=CLASSIFICATION if is_classification else REGRESSION)
report = da.dataset_analysis_report(dataset=train)
return report