Upload app.py

app.py

@@ -25,6 +25,7 @@ import gradio as gr
 import random
 import re
 
+
 #Read data.
 from datasets import load_dataset
 x1 = load_dataset("mertkarabacak/NTDB-Epidural", data_files="mortality_data.csv", use_auth_token = HF_TOKEN)
@@ -32,6 +33,7 @@ x1 = pd.DataFrame(x1['train'])
 variables = ['Age', 'Sex', 'Ethnicity', 'Weight', 'Height', 'Systolic_Blood_Pressure', 'Pulse_Rate', 'Supplemental_Oxygen', 'Pulse_Oximetry', 'Respiratory_Assistance', 'Respiratory_Rate', 'Temperature', 'PreHospital_Cardiac_Arrest', 'GCS__Eye', 'GCS__Verbal', 'GCS__Motor', 'Total_GCS', 'Pupillary_Response', 'Midline_Shift', 'Current_Smoker', 'Comorbid_Condition__Alcohol_Use_Disorder', 'Comorbid_Condition__Substance_Abuse_Disorder', 'Comorbid_Condition__Diabetes_Mellitus', 'Comorbid_Condition__Hypertension', 'Comorbid_Condition__Congestive_Heart_Failure', 'History_of_Myocardial_Infarction', 'Comorbid_Condition__Angina_Pectoris', 'History_of_Cerebrovascular_Accident', 'Comorbid_Condition__Peripheral_Arterial_Disease', 'Comorbid_Condition__Chronic_Obstructive_Pulmonary_Disease', 'Comorbid_Condition__Chronic_Renal_Failure', 'Comorbid_Condition__Cirrhosis', 'Comorbid_Condition__Bleeding_Disorder', 'Comorbid_Condition__Disseminated_Cancer', 'Comorbid_Condition__Currently_Receiving_Chemotherapy_for_Cancer', 'Comorbid_Condition__Dementia', 'Comorbid_Condition__Attention_Deficit_Disorder_or_Attention_Deficit_Hyperactivity_Disorder', 'Comorbid_Condition__Mental_or_Personality_Disorder', 'Ability_to_Complete_AgeAppropriate_ADL', 'Pregnancy', 'Anticoagulant_Therapy', 'Steroid_Use', 'Advanced_Directive_Limiting_Care', 'Days_from_Incident_to_ED_or_Hospital_Arrival', 'Transport_Mode', 'InterFacility_Transfer', 'Trauma_Type', 'Injury_Intent', 'Mechanism_of_Injury', 'WorkRelated', 'AIS_Severity__Maximum_Severity_of_Injury_in_the_Head', 'AIS_Severity__Maximum_Severity_of_Injury_in_the_Face', 'AIS_Severity__Maximum_Severity_of_Injury_in_the_Neck', 'AIS_Severity__Maximum_Severity_of_Injury_in_the_Thorax', 'AIS_Severity__Maximum_Severity_of_Injury_in_the_Abdomen', 'AIS_Severity__Maximum_Severity_of_Injury_in_the_Spine', 'AIS_Severity__Maximum_Severity_of_Injury_in_the_Upper_Extremity', 'AIS_Severity__Maximum_Severity_of_Injury_in_the_Lower_Extremity', 'AIS_Severity__Maximum_Severity_of_Injury_in_Unspecified_Body_Regions', 'AIS_derived_ISS', 'Blood_Transfusion', 'Neurosurgical_Intervention', 'Alcohol_Screen', 'Alcohol_Screen_Result', 'Drug_Screen__Amphetamine', 'Drug_Screen__Barbiturate', 'Drug_Screen__Benzodiazepines', 'Drug_Screen__Cannabinoid', 'Drug_Screen__Cocaine', 'Drug_Screen__MDMA_or_Ecstasy', 'Drug_Screen__Methadone', 'Drug_Screen__Methamphetamine', 'Drug_Screen__Opioid', 'Drug_Screen__Oxycodone', 'Drug_Screen__Phencyclidine', 'Drug_Screen__Tricyclic_Antidepressant', 'ACS_Verification_Level', 'Hospital_Type', 'Facility_Bed_Size', 'Primary_Method_of_Payment', 'Race', 'Cerebral_Monitoring', 'Protective_Device', 'OUTCOME']
 x1 = x1[variables]
 
+
 #Assign unique values as answer options.
 unique_SEX = ['Male', 'Female', 'Unknown']
 unique_RACE = ['White', 'Black', 'Asian', 'American Indian', 'Pacific Islander', 'Other', 'Unknown']
@@ -93,6 +95,7 @@ unique_HOSPITALTYPE = ['Non-profit', 'For profit', 'Government', 'Unknown']
 unique_BEDSIZE = ['More than 600', '401 to 600', '201 to 400', '200 or fewer']
 unique_PRIMARYMETHODPAYMENT = ['Private/commercial insurance', 'Medicaid', 'Medicare', 'Other government', 'Self-pay',  'Other', 'Not billed', 'Unknown']
 
+
 #Prepare data for the outcome 1 (mortality).
 y1 = x1.pop('OUTCOME')
 resampler = SMOTE(random_state = 0)
@@ -112,11 +115,13 @@ for col in categorical_columns1:
 d1 = dict.fromkeys(x1_rf.select_dtypes(np.int64).columns, str)
 x1_rf = x1_rf.astype(d1)
 
+
 #Assign hyperparameters.
-y1_xgb_params = {'objective': 'binary:logistic', 'booster': 'gbtree', 'lambda': 2.0550082441612586e-08, 'alpha': 2.9754583561289483e-06, 'max_depth': 5, 'eta': 6.610855432422906e-07, 'gamma': 6.373311989945556e-07, 'grow_policy': 'lossguide'}
-y1_lgb_params = {'objective': 'binary', 'boosting_type': 'gbdt', 'lambda_l1': 0.19483841111945205, 'lambda_l2': 0.00042014242263296316, 'num_leaves': 2, 'feature_fraction': 0.7765322549174258, 'bagging_fraction': 0.6255184223663047, 'bagging_freq': 6, 'min_child_samples': 80}
-y1_cb_params = {'objective': 'CrossEntropy', 'colsample_bylevel': 0.02048734243685444, 'depth': 9, 'boosting_type': 'Ordered', 'bootstrap_type': 'MVS'}
-y1_rf_params = {'criterion': 'gini', 'max_features': 'log2', 'max_depth': 1, 'n_estimators': 700, 'min_samples_leaf': 3, 'min_samples_split': 4}
+y1_xgb_params = {'objective': 'binary:logistic', 'booster': 'gbtree', 'lambda': 3.540855010579091e-08, 'alpha': 4.005546508605542e-08, 'max_depth': 5, 'eta': 5.190362998186933e-08, 'gamma': 1.1458984717217304e-05, 'grow_policy': 'depthwise'}
+y1_lgb_params = {'objective': 'binary', 'boosting_type': 'gbdt', 'lambda_l1': 0.5873889067286373, 'lambda_l2': 0.0043364331356120405, 'num_leaves': 221, 'feature_fraction': 0.8036023624154648, 'bagging_fraction': 0.43028542431491096, 'bagging_freq': 5, 'min_child_samples': 7}
+y1_cb_params = {'objective': 'Logloss', 'colsample_bylevel': 0.02893626365611705, 'depth': 9, 'boosting_type': 'Plain', 'bootstrap_type': 'Bayesian', 'bagging_temperature': 0.8396785798333539}
+y1_rf_params = {'criterion': 'entropy', 'max_features': None, 'max_depth': 71, 'n_estimators': 900, 'min_samples_leaf': 4, 'min_samples_split': 5}
+
 
 #Modeling for y1 (mortality).
 y1_model_xgb = xgb.train(params=y1_xgb_params, dtrain=y1_data_xgb)
@@ -133,6 +138,7 @@ y1_rf = rf(**y1_rf_params)
 y1_model_rf = y1_rf.fit(x1_rf, y1)
 y1_explainer_rf = shap.TreeExplainer(y1_model_rf)
 
+
 #Define predict for y1 (mortality).
 def y1_predict_xgb(*args):
     df_xgb = pd.DataFrame([args], columns=x1.columns)
@@ -222,6 +228,7 @@ def y1_interpret_rf(*args):
     plt.ylabel("Feature", fontsize = 16, labelpad=14, fontweight = 'bold')
     return fig_m
 
+
 with gr.Blocks(title = "NTDB-Epidural") as demo:
     
     gr.Markdown(