def test_neural_network_basic():
runner = Runner('model/experiment/output/neural_network_basic',
load_sample_data_frame(), 'violation',
neural_network_basic, None)
runner.run_classification_experiment(sample=sample,
multiclass=True,
record_predict_proba=True)
def test_gaussian_naive_bayes_basic():
runner = Runner('model/experiment/output/complement_naive_bayes_basic',
load_sample_data_frame(), 'violation',
gaussian_naive_bayes_basic, None)
runner.run_classification_experiment(sample=sample,
multiclass=True,
record_predict_proba=True)
def build_xgboost_model():
runner = Runner('model/output/xgboost_model', load_data_frame(),
'violation', xgboost_pipeline, None)
runner.run_classification_experiment(sample=sample,
test_size=0.2,
multiclass=True,
record_predict_proba=True)
joblib.dump(xgboost_pipeline, 'model/output/xgboost_model.joblib')
def build_sgd_huber_loss():
runner = Runner(
'model/output/sgd_huber_loss_over_sampled',
load_clean_data_frame(),
'arrest',
sgd
)
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
sampling=SMOTE(),
fit_increment=fit_increment,
max_iters=max_iters,
n_jobs=1,
cv=None
)
joblib.dump(
pipeline,
'model/output/sgd_huber_loss_over_sampled.joblib'
)
runner = Runner(
'model/output/sgd_huber_loss_over_sampled_fs',
load_clean_data_frame(),
'arrest',
sgd_fs
)
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=mapper,
sampling=SMOTE(),
fit_increment=fit_increment,
max_iters=max_iters,
n_jobs=1,
cv=None
)
joblib.dump(
pipeline_fs,
'model/output/sgd_huber_loss_over_sampled_fs.joblib'
)
def build_neural_network():
runner = Runner('model/output/neural_network_basic',
load_clean_data_frame(), 'arrest', nn)
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
fit_increment=fit_increment,
max_iters=max_iters,
cv=None,
n_jobs=1)
joblib.dump(pipeline, 'model/output/neural_network_basic.joblib')
runner = Runner('model/output/neural_network_basic_fs',
load_clean_data_frame(), 'arrest', nn_fs)
runner.run_classification_experiment(sample=sample,
record_predict_proba=True,
transformer=mapper,
fit_increment=fit_increment,
max_iters=max_iters,
cv=None,
n_jobs=1)
joblib.dump(pipeline_fs, 'model/output/neural_network_basic_fs.joblib')
def test_sgd_log_loss():
runner = Runner('model/experiment/output/sgd_log_loss_basic',
load_clean_sample_data_frame(), 'arrest',
SGDClassifier(loss='log'))
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
fit_increment=fit_increment,
max_iters=max_iters,
n_jobs=1)
runner = Runner('model/experiment/output/sgd_log_loss_under_sampled',
load_clean_sample_data_frame(), 'arrest',
SGDClassifier(loss='log'))
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
fit_increment=fit_increment,
max_iters=max_iters,
n_jobs=1,
sampling=RandomUnderSampler())
runner = Runner('model/experiment/output/sgd_log_loss_over_sampled',
load_clean_sample_data_frame(), 'arrest',
SGDClassifier(loss='log'))
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
fit_increment=fit_increment,
max_iters=max_iters,
n_jobs=1,
sampling=SMOTE())
runner = Runner('model/experiment/output/sgd_log_loss_combine_sampled',
load_clean_sample_data_frame(), 'arrest',
SGDClassifier(loss='log'))
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
fit_increment=fit_increment,
max_iters=max_iters,
n_jobs=1,
sampling=SMOTEENN())
def test_gaussian_naive_bayes():
runner = Runner('model/experiment/output/gaussian_naive_bayes_basic',
load_clean_sample_data_frame(), 'arrest', GaussianNB())
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
fit_increment=fit_increment,
n_jobs=1)
runner = Runner(
'model/experiment/output/gaussian_naive_bayes_under_sampled',
load_clean_sample_data_frame(), 'arrest', GaussianNB())
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
fit_increment=fit_increment,
n_jobs=1,
sampling=RandomUnderSampler())
runner = Runner(
'model/experiment/output/gaussian_naive_bayes_over_sampled',
load_clean_sample_data_frame(), 'arrest', GaussianNB())
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
fit_increment=fit_increment,
n_jobs=1,
sampling=SMOTE())
runner = Runner(
'model/experiment/output/gaussian_naive_bayes_combine_sampled',
load_clean_sample_data_frame(), 'arrest', GaussianNB())
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
fit_increment=fit_increment,
n_jobs=1,
sampling=SMOTEENN())
def test_neural_network():
runner = Runner(
'model/experiment/output/neural_network_basic',
load_clean_sample_data_frame(), 'arrest',
MLPClassifier(hidden_layer_sizes=(
750,
125,
), verbose=True))
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
fit_increment=fit_increment,
max_iters=max_iters,
n_jobs=1)
runner = Runner(
'model/experiment/output/neural_network_under_sampled',
load_clean_sample_data_frame(), 'arrest',
MLPClassifier(hidden_layer_sizes=(
750,
125,
), verbose=True))
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
fit_increment=fit_increment,
max_iters=max_iters,
n_jobs=1,
sampling=RandomUnderSampler())
runner = Runner(
'model/experiment/output/neural_network_over_sampled',
load_clean_sample_data_frame(), 'arrest',
MLPClassifier(hidden_layer_sizes=(
750,
125,
), verbose=True))
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
fit_increment=fit_increment,
max_iters=max_iters,
n_jobs=1,
sampling=SMOTE())
runner = Runner(
'model/experiment/output/neural_network_combine_sampled',
load_clean_sample_data_frame(), 'arrest',
MLPClassifier(hidden_layer_sizes=(
750,
125,
), verbose=True))
runner.run_classification_experiment(
sample=sample,
record_predict_proba=True,
transformer=binned_geo_one_hot_data_mapper,
fit_increment=fit_increment,
max_iters=max_iters,
n_jobs=1,
sampling=SMOTEENN())
