import analyzer.dataset as ds
import evaluation.treatment_utils as u
#%% Choose variant and specify paths
treatment = 'ACEI_ARBS'
outcome = 'COMORB_DEATH'
path = '../../covid19_treatments_results/matched_single_treatments_hypertension/' + treatment
website_path = '../../website/assets/treatment_calculators/' + outcome
training_set_name = treatment + '_hope_hm_cremona_matched_all_treatments_train.csv'
X, Z, y = u.load_data(
'../../covid19_treatments_data/matched_single_treatments_hypertension/',
training_set_name,
split='train',
matched=True,
prediction=outcome,
other_tx=False,
med_hx=False)
X.drop('HYPERTENSION', axis=1, inplace=True)
load_file_path_yes = os.path.join(path, outcome, 'cart',
'ACEI_ARBS_matched_comorb_death_seed1')
with open(load_file_path_yes, 'rb') as file:
load_file_yes = pickle.load(file)
X_test_yes = load_file_yes['train'].drop(outcome, axis=1)
load_file_path_no = os.path.join(path, outcome, 'lr',
training_set_name = treatment + train_file
results_path = '../../covid19_treatments_results/'
version_folder = version + str(treatment) + '/' + str(outcome) + '/'
save_path = results_path + version_folder + 'summary/'
training_set_name = treatment + train_file
data_version = 'train'
threshold = 0.05
train_X, Z, y = u.load_data(data_path,
training_set_name,
split=data_version,
matched=matched,
prediction=outcome,
other_tx=False,
replace_na='NO_' + treatment)
df_result = pd.read_csv(save_path + data_version + '_' + match_status +
'_bypatient_allmethods_benefit.csv')
benefit = df_result.groupby('ID').agg({'Benefit': 'mean'})['Benefit']
### Run Model
grid = iai.GridSearch(
iai.OptimalTreeRegressor(random_seed=1, ),
max_depth=range(3, 8),
)
grid.fit(X, benefit)
metrics_agg = pd.DataFrame(columns = ['data_version','weighted_status','threshold','match_rate','presc_count','average_auc',
'PE_0','CPE_0','PE','CPE','pr_low','pr_high'])
for outcome in prediction_list:
version_folder = str(treatment)+'/'+str(outcome)+'/'
save_path = results_path + version_folder + 'summary/'
# create summary folder if it does not exist
Path(save_path).mkdir(parents=True, exist_ok=True)
for data_version in data_list:
print(data_version)
for threshold in [0,0.01,0.02,0.05,0.1]:
print('Threshold = '+str(threshold))
#Read in the relevant data
X, Z, y = u.load_data(data_path,training_set_name,
split=data_version, matched=matched, prediction = outcome,
replace_na = 'NO_'+treatment)
result = pd.read_csv(save_path+data_version+'_'+match_status+'_bypatient_allmethods.csv')
#Filter only to algorithms in the algorithms list
result =result.loc[result['Algorithm'].isin(algorithm_list)]
# result.set_index(['ID','Algorithm'], inplace = True)
pred_results = pd.read_csv(save_path+data_version+'_'+match_status+'_performance_allmethods.csv')
pred_results.set_index('Algorithm', inplace = True)
#Predictive performance table to base decisions from
pred_perf_results = pd.read_csv(save_path+'train'+'_'+match_status+'_performance_allmethods.csv')
pred_perf_results.set_index('Algorithm', inplace = True)
#Compare different schemes
treatment = 'CORTICOSTEROIDS'
treatment_list = [treatment, 'NO_'+treatment]
results_path = '../../covid19_treatments_results/'
version_folder = 'matched_single_treatments_der_val_addl_outcomes/'+str(treatment)+'/'+str(outcome)+'/'
save_path = results_path + version_folder + 'summary/'
training_set_name = treatment+'_hope_hm_cremona_matched_all_treatments_train.csv'
#%% Run results
if not preload:
# create summary folder if it does not exist
Path(save_path).mkdir(parents=True, exist_ok=True)
for data_version in ['train','test','validation','validation_cremona','validation_hope','validation_hope_italy']:
print(data_version)
X, Z, y = u.load_data(data_path,training_set_name,
split=data_version, matched=matched, prediction = outcome)
print("X observations: "
, str(X.shape[0]))
result = pd.concat([u.algorithm_predictions(X, treatment_list = treatment_list,
algorithm = alg, matched = matched,
prediction = outcome,
result_path = results_path+version_folder)
for alg in algorithm_list], axis = 0)
# Find optimal prescription across methods
result['Prescribe'] = result.idxmin(axis=1)
result['Prescribe_Prediction'] = result.min(axis=1)
# Save result file
result.to_csv(save_path+data_version+'_'+match_status+'_bypatient_allmethods.csv')
# =============================================================================
# Predictive Performance evaluation:
# - Given a combination of treatment and method calculate the AUC
algorithm_list = ['rf', 'cart', 'qda', 'gb', 'xgboost']
# data_list = ['train','test']
treatment = 'ACEI_ARBS'
treatment_list = [treatment, 'NO_' + treatment]
version_folder = str(treatment) + '/' + str(outcome) + '/'
save_path = results_path + version_folder + 'summary/'
## Load data for comparison
training_set_name = treatment + '_hope_hm_cremona_matched_all_treatments_train.csv'
X_train, Z_train, y_train = u.load_data(data_path + version,
training_set_name,
split='train',
matched=matched,
prediction=outcome,
med_hx=False,
other_tx=False)
#%% Match new data to training format
## Impute missing data
data_oh = pd.get_dummies(data, prefix_sep='_', drop_first=True)
X = data_oh.reindex(X_train.columns, axis=1)
Z = data[treatment].apply(lambda x: treatment
if x == 1 else 'NO_' + treatment).rename('REGIMEN')
y = data[outcome]
ft_val = X.describe().transpose().add_prefix('val_')
ft_train = X_train.describe().transpose().add_prefix('train_')