def test_treefoo_duh(self):
bundle, datasets, stations_df = bltu.make_basic_minimal_model()
holdout_df = datasets['holdout_df']
y_predictions, y_test, metrics = blc.run_model_predict(bundle,
holdout_df,
stations_df,
labeled=True)
proportion_correct_labeled_true = bltu.get_basic_proportion_correct(
y_test, y_predictions)
# Again but unlabeled now.
contracted_df = blc.contract_df(holdout_df)
widened_df = blc.widen_df_with_other_cols(contracted_df, s.ALL_COLUMNS)
y_predictions_from_widened, _, _ = blc.run_model_predict(bundle,
widened_df,
stations_df,
labeled=False)
proportion_correct_labeled_false = bltu.get_basic_proportion_correct(
y_test, y_predictions_from_widened)
# and assert, evaluation should be the same..
assert proportion_correct_labeled_false == proportion_correct_labeled_true
pass
def test_basic(self):
csvdata = '10/1/2015 00:00:02,W 26 St & 10 Ave,Subscriber,1973,1\n10/1/2015 00:00:02,E 39 St & 2 Ave,Subscriber,1990,1'
df = blc.hydrate_and_widen(csvdata)
bundle, datasets, stations_df = bltu.make_basic_minimal_model()
y_predictions, _, _ = blc.run_model_predict(bundle,
df,
stations_df,
labeled=False)
pass
def test_assertion_happens_for_nan(self):
_, datasets, stations_df = bltu.make_basic_minimal_model()
df = pd.DataFrame({s.NEW_END_NEIGHBORHOOD: ['foo', 'nan']})
feature_encoding_dict = {s.NEW_END_NEIGHBORHOOD: str}
asserted = False
try:
_, label_encoders = pl.make_simple_df_from_raw(
df, stations_df, feature_encoding_dict)
except Exception:
asserted = True
assert asserted
def test_treefoo_with_pure_input_data(self):
# csvdata = 'starttime,start station name,usertype,birth year,gender\n10/1/2015 00:00:02,W 26 St & 10 Ave,Subscriber,1973,1\n10/1/2015 00:00:02,E 39 St & 2 Ave,Subscriber,1990,1'
csvdata = '10/1/2015 00:00:02,W 26 St & 10 Ave,Subscriber,1973,1\n10/1/2015 00:00:02,E 39 St & 2 Ave,Subscriber,1990,1'
df = blc.hydrate_csv_to_df(csvdata)
bundle, datasets, stations_df = bltu.make_basic_minimal_model()
widened_df = blc.widen_df_with_other_cols(df, s.ALL_COLUMNS)
y_predictions, _, _ = blc.run_model_predict(bundle,
widened_df,
stations_df,
labeled=False)
pass
def test_basic(self):
from nose.tools import set_trace
set_trace()
csvdata = '10/1/2015 00:00:02,E 39 St & 2 Ave,Subscriber,1990,1'
df = blc.hydrate_and_widen(csvdata)
bundle, datasets, stations_df = bltu.make_basic_minimal_model()
_, X_test = blc.df_to_np_for_clf(bundle,
df,
stations_df,
labeled=False)
clf = bundle['clf']
path = clf.decision_path(X_test)
def test_predict_matches_proba(self):
bundle, datasets, stations_df = bltu.make_basic_minimal_model()
df = datasets['holdout_df']
clf = bundle['clf']
prepared = blc.predict_prepare(bundle, df, stations_df, labeled=True)
y_topk1_outputs = blmu.get_sorted_predict_proba_predictions(
prepared['y_predict_proba'], clf.classes_, k=1)
y_predictions, y_test, metrics = blc.run_model_predict(bundle,
df,
stations_df,
labeled=True)
l = [
x[1] for x in sorted(metrics['rank_k_proba_scores'].items(),
key=lambda x: x[0])
]
assert all([l[i] <= l[i + 1] for i in range(len(l) - 1)])
assert [x[0] for x in y_topk1_outputs] == list(y_predictions)