def test_combine_selector_on_normal_cases(self):
a = []
b = ['add_numeric', 'multiply_numeric']
x = FeatureEng('file_1.csv', 'infer', None, 'target', a, b, 1, 0)
x.new_features()
x_2 = FeatureSelect(x.feature_matrix, 6)
x_2.combine_selector()
#define the correct result dataframe that we expect to get
values = [['feat3', True, True, True, True, True, True, 6],
['feat2', True, True, True, True, True, True, 6],
['feat1 + feat3', True, True, True, True, True, True, 6],
['feat1 + feat2', True, True, True, True, True, True, 6],
['feat2 * feat3', True, True, True, False, False, False, 3],
['feat1 * feat3', True, True, True, False, False, False, 3]]
col_names = [
'Feature', 'Pearson', 'Chi-2', 'Logistics', 'Random Forest',
'LightGBM', 'Extra_trees', 'Total'
]
df_result = pd.DataFrame(data=values, columns=col_names)
df_result.index += 1
assert_frame_equal(df_result, x_2.combine_selector())
def test_df_with_new_features_on_normal_values(self):
a=[]
b=['add_numeric', 'multiply_numeric']
x=FeatureEng('file_1.csv','infer', None,'target',a,b,2,0.8)
feat_matrix = x.df_with_new_features()
message = "column names do not match"
correct_columns = ['feat1',
'feat2',
'feat3',
'feat1 + feat3',
'feat1 + feat2',
'feat2 + feat3',
'feat1 * feat3',
'feat2 * feat3',
'feat1 * feat2',
'feat1 + feat2 * feat2 + feat3',
'feat1 + feat3 * feat2 + feat3',
'feat1 + feat2 * feat3',
'feat1 + feat2 * feat1 + feat3',
'feat1 + feat2 * feat2',
'feat1 * feat2 + feat3',
'feat1 * feat1 + feat2',
'feat1 + feat3 * feat2',
'feat1 + feat3 * feat3',
'feat2 + feat3 * feat3',
'feat1 * feat1 + feat3',
'feat2 * feat2 + feat3',
'target']
assert feat_matrix.columns.isin(correct_columns).all(), message
assert not ~ feat_matrix.columns.isin(correct_columns).all(), message
def test_new_features_on_containing_non_numeric_values(self):
with pytest.raises(ValueError) as exc_info:
a = []
b = ['add_numeric', 'multiply_numeric']
x = FeatureEng("file_2.csv", 'infer', None, 'target', a, b, 2, 0.8)
x.new_features()
assert exc_info.match("Data Frame contains non-numeric values")
def test_df_with_new_features_on_containing_target_column(self):
#test when target column exists in dataset (it was not removed correctly by separate_features_from_label stage)
with pytest.raises(ValueError) as exc_info_2:
a = []
b = ['add_numeric', 'multiply_numeric']
x = FeatureEng('file_1.csv', 'infer', None, 'feat1', a, b, 2, 0.8)
x.df_with_new_features()
assert exc_info_2.match("Data Frame already contains target column")
def test_df_selected_columns_normal_case(self):
a=[]
b=['add_numeric', 'multiply_numeric']
x = FeatureEng('file_1.csv','infer',None,'target',a,b,4,0.5)
x.df_with_new_features()
x_2 = FeatureSelect(x.df_new_features, 3)
x_2.save_df_selected_columns()
#define the correct result dataframe that we expect to get
values = [[2,5,5,0],
[5,5,5,1],
]
col_names = ['feat2','feat2 + feat3','feat1 * feat2 + feat3','target']
df_result = pd.DataFrame(data = values, columns = col_names)
#df_result.index += 0
x_2.df_selected_columns.reset_index(drop=True, inplace=True)
df_result.reset_index(drop=True, inplace=True)
assert_frame_equal(df_result,x_2.df_selected_columns)