def build_auto(regressor, name, **pmml_options):
cat_columns = ["cylinders", "model_year", "origin"]
cont_columns = ["displacement", "horsepower", "weight", "acceleration"]
if isinstance(regressor, LGBMRegressor):
cat_mappings = [([cat_column], [cat_domain(name), label_encoder(name)]) for cat_column in cat_columns]
else:
cat_mappings = [([cat_column], [cat_domain(name), label_binarizer(name)]) for cat_column in cat_columns]
cont_mappings = [([cont_column], [cont_domain(name)]) for cont_column in cont_columns]
mapper = DataFrameMapper(cat_mappings + cont_mappings)
pipeline = PMMLPipeline([
("mapper", mapper),
("regressor", regressor)
])
if isinstance(regressor, LGBMRegressor):
pipeline.fit(auto_X, auto_y, regressor__categorical_feature = [0, 1, 2])
elif isinstance(regressor, IsolationForest):
pipeline.fit(auto_X)
else:
pipeline.fit(auto_X, auto_y)
if isinstance(regressor, XGBRegressor):
pipeline.verify(auto_X.sample(n = 3, random_state = 13), precision = 1e-5, zeroThreshold = 1e-5)
else:
pipeline.verify(auto_X.sample(n = 3, random_state = 13))
pipeline.configure(**pmml_options)
store_pmml(pipeline, name)
if isinstance(regressor, IsolationForest):
decision_function = DataFrame(pipeline.decision_function(auto_X), columns = ["decisionFunction"])
outlier = DataFrame(pipeline.predict(auto_X), columns = ["outlier"])
outlier['outlier'] = outlier['outlier'].apply(lambda x: str(bool(x == -1)).lower())
store_csv(pandas.concat((decision_function, outlier), axis = 1), name)
else:
mpg = DataFrame(pipeline.predict(auto_X), columns = ["mpg"])
store_csv(mpg, name)
])
# Train the model
pipeline.fit(train_data, train_labels)
# Pack verification data and verify
pipeline.verify(test_data)
# Save the pipeline + model
sklearn2pmml.sklearn2pmml(sklearn2pmml.make_pmml_pipeline(
pipeline, active_fields=FEATURE_COLS, target_fields=TARGET_COL),
'model.pmml',
with_repr=True,
debug=True)
# Measure the accuracy
descision = pipeline.decision_function(test_data.sample(1))
jpmml_input_data = pd.read_csv("jpmml-test-input.csv", usecols=FEATURE_COLS)
input_pred = pipeline.predict(jpmml_input_data.squeeze())
with numpy.printoptions(threshold=numpy.inf):
print(input_pred)
pred = pipeline.predict(test_data)
print('Accuracy = {:.3f}'.format(
sum(l == p for l, p in zip(test_labels, pred)) / len(test_labels)))
print(
'Dominant label freq = {:.3f}'.format(1 -
sum(test_labels) / len(test_labels)))
print('ROC AUC = {:.3f}'.format(roc_auc_score(test_labels, pred)))
print(len(train_labels), 'training data rows')
print(len(pipeline[1].coef_[0]), 'model parameters')
