def collect_results(dataset,
q_codes,
results: Dict[str, List[float]],
identifier='weka',
q_idx: List[int] = None,
**extra_fields):
# Init
df = pd.DataFrame()
nb_qry, n_att = q_codes.shape
miss_encoding = encode_attribute(2, [0], [1])
n_miss = np.sum((q_codes == miss_encoding), axis=1)
# Build DataFrame
df['missing_percentage'] = n_miss / n_att
df['difficulty'] = _convert_percentage_missing_to_difficulty(
df['missing_percentage'].values)
if q_idx is not None:
df['q_idx'] = q_idx
else:
df['q_idx'] = range(nb_qry)
for k in results:
df[k] = results[k]
df['identifier'] = identifier
df['dataset'] = dataset
for f in extra_fields:
df[f] = extra_fields[f]
return df
def setup_classification():
train, test = datasets.load_nursery()
model = MERCS()
ind_parameters = {
'ind_type': 'RF',
'ind_n_estimators': 10,
'ind_max_depth': 4
}
sel_parameters = {'sel_type': 'Base', 'sel_its': 4, 'sel_param': 2}
model.fit(train, **ind_parameters, **sel_parameters)
code = [0, 0, 0, 0, 0, 0, 0, 0, 1]
target_boolean = np.array(code) == encode_attribute(2, [1], [2])
y_true = test[test.columns.values[target_boolean]].values
return train, test, code, model, y_true
def test_perform_imputation():
# Prelims
train, test = ds.load_nursery()
query_code = np.array([0, -1, -1, -1, -1, -1, 0, 0, 1])
imputator = Imputer(missing_values='NaN',
strategy='most_frequent',
axis=0)
imputator.fit(train)
# Actual test
obs = perform_imputation(test, query_code, imputator)
assert test.shape == obs.shape
assert isinstance(obs, np.ndarray)
boolean_missing = encode_attribute(0, [1], [2])
for row in obs[:, boolean_missing].T:
assert len(np.unique(row)) == 1
def setup():
train, test = datasets.load_fertility()
model = MERCS()
# Ensure attributes are correctly recognized as nominal/numeric
train['season'] = pd.factorize(train['season'])[0]
test['season'] = pd.factorize(test['season'])[0]
ind_parameters = {'ind_type': 'RF',
'ind_n_estimators': 10,
'ind_max_depth': 4}
sel_parameters = {'sel_type': 'Base',
'sel_its': 8,
'sel_param': 2}
model.fit(train, **ind_parameters, **sel_parameters)
code = [0, 0, 0, 0, 0, 0, 0, 0, 1, 1]
target_boolean = np.array(code) == encode_attribute(2, [1], [2])
y_true = test[test.columns.values[target_boolean]].values
return train, test, code, model, y_true, target_boolean