def test_unique_values(self):
df = pd.read_csv(fixture('iris_classification.csv'), na_values=['None'])
unique_drgs = len(df.DRG.unique())
test_df = impact_coding_on_a_single_column(df, 'species', 'DRG')
unique_impact_values = len(test_df.DRG_impact_coded.unique())
self.assertLessEqual(unique_impact_values, unique_drgs)
def test_column_renaming(self):
df = pd.read_csv(fixture('iris_classification.csv'),
na_values=['None'])
code_column_name = 'DRG'
test_df = impact_coding_on_a_single_column(df, 'species',
code_column_name)
self.assertTrue((code_column_name +
'_impact_coded') in test_df.columns)
def setUp(self):
cols = ['ThirtyDayReadmitFLG', 'SystolicBPNBR', 'LDLNBR']
df = pd.read_csv(fixture('HCPyDiabetesClinical.csv'),
na_values=['None'],
usecols=cols)
np.random.seed(42)
self.o = DevelopSupervisedModel(modeltype='classification',
df=df,
predictedcol='ThirtyDayReadmitFLG',
impute=True)
def setUp(self):
df = pd.read_csv(fixture('HCPyDiabetesClinical.csv'),
na_values=['None'])
# Drop uninformative columns
df.drop(['PatientID', 'InTestWindowFLG'], axis=1, inplace=True)
np.random.seed(42)
self.o = DevelopSupervisedModel(modeltype='classification',
df=df,
predictedcol='ThirtyDayReadmitFLG',
impute=True)
self.o.linear(cores=1)
def setUp(self):
df = pd.read_csv(fixture('DiabetesClinicalSampleData.csv'),
na_values=['None'])
# Drop uninformative columns
df.drop(['PatientID', 'InTestWindowFLG'], axis=1, inplace=True)
# Convert numeric columns to factor/category columns
np.random.seed(42)
self.o = DevelopSupervisedModel(modeltype='classification',
df=df,
predictedcol='ThirtyDayReadmitFLG',
impute=True)
self.o.random_forest(cores=1)
def setUp(self):
df = pd.read_csv(fixture('HCPyDiabetesClinical.csv'),
na_values=['None'])
df.drop('PatientID', axis=1, inplace=True) # drop uninformative column
np.random.seed(42)
self.o = DeploySupervisedModel(modeltype='classification',
df=df,
graincol='PatientEncounterID',
windowcol='InTestWindowFLG',
predictedcol='ThirtyDayReadmitFLG',
impute=True)
self.o.deploy(
method='linear',
cores=1,
server='localhost',
dest_db_schema_table='[SAM].[dbo].[HCPyDeployClassificationBASE]',
use_saved_model=False)
