def load_assess_files(segment):
if segment["model"] != 'Logistic Regression':
cols = segment['session'].CASTable(segment['score_params']['casOut']['name']).columns
df = segment['session'].CASTable(segment['score_params']['casOut']['name'])[[cols[0],cols[-2]]].to_frame()
df.dropna(inplace=True)
yTestActual = df[cols[0]].tolist()
yTestPredict = df[cols[-2]].tolist()
data = [(None, None),
(None, None),
(yTestActual, yTestPredict)]
# Define a directory where all the files for model manager will be kept
MODEL_DIR = 'automate/{}_{}-{}'.format(segment['model'], segment['region'], segment['handset'])
if os.path.exists(MODEL_DIR):
shutil.rmtree(MODEL_DIR)
os.makedirs(MODEL_DIR)
JSONFiles = pzmm.JSONFiles()
target = cols[0]
JSONFiles.calculateFitStat(data, MODEL_DIR)
JSONFiles.generateROCStat(data, target, MODEL_DIR)
JSONFiles.generateLiftStat(data, target, 1, MODEL_DIR)
print('Model exists, creting new version')
model_repository.delete_model_contents(modelname)
register_model(model=model,
name=modelname,
project=project,
input=inputs,
force=True,
version='latest')
### adding extra files
### not needed but good practice
path = Path.cwd()
####
#### Creating input & output files
JSONFiles = pzmm.JSONFiles()
### write inputVar.json
JSONFiles.writeVarJSON(inputs, isInput=True, jPath=path)
### write outputVar.json
JSONFiles.writeVarJSON(outputs, isInput=False, jPath=path)
#### missing files files
filenames = {
'file': ['inputVar.json', 'outputVar.json', 'training_code_iris.py'],
'role': ['input', 'output', 'train']
}
#### uploading files
def score_non_swat_model(segment, tab, authUser):
df = segment['segment_tbl'].to_frame()
df = df[df['_PartInd_'] == 0]
df = df.select_dtypes(exclude=['object'])
imputed_cols = [col for col in df.columns if "IMP" in col]
df = df[imputed_cols]
target = "IMP_" + tab.children[0].children[3].children[0].children[0].value
imputed_cols.remove(target)
X_test = df[imputed_cols]
y_test = df[target].tolist()
if segment['model'] =='XGBoost':
y_predicted = segment['modelObj'].predict_proba(X_test)[:,1]
fpr, tpr, _ = roc_curve(y_test, y_predicted)
ROCInfo = dict(FPR = fpr, Sensitivity = tpr)
segment['ROCInfo'] = ROCInfo
segment['LIFTInfo'] = create_lift(y_test, y_predicted)
accuracy = accuracy_score(y_test, segment['modelObj'].predict(X_test))
segment['misclassification'] = 1 - accuracy
elif segment['model'] =='TensorFlow':
#X_test = (X_test - X_test.mean()) / (X_test.max() - X_test.min())
y_predicted = segment['modelObj'].predict(X_test.as_matrix()).flatten()
fpr, tpr, _ = roc_curve(y_test, y_predicted)
ROCInfo = dict(FPR = fpr, Sensitivity = tpr)
segment['ROCInfo'] = ROCInfo
segment['LIFTInfo'] = create_lift(y_test, y_predicted)
accuracy = accuracy_score(y_test, [round(num) for num in y_predicted])
segment['misclassification'] = 1 - accuracy
elif (segment['model'] =='AutoML') and (segment['modelObj'] is not None):
segment['session'].loadactionset('modelpublishing')
tbl_test = segment['segment_tbl'].query('_PartInd_ = 0')
y_test = tbl_test[tab.children[0].children[3].children[0].children[0].value].values.tolist()
var_list = list()
for col in list(segment['session'].CASTable(tab.children[0].children[1].value, caslib='Public').columns):
var_list.append({"name" : col})
model_tbl = segment['session'].loadtable(path='sas_model_table.sashdat',caslib='Public', casout={"caslib" : "CASUSER({})".format(authUser)})
segment['session'].table.partition(table={'name' : tbl_test.name,
"vars" : var_list, "where" : "_PartInd_ = 0"},
casout={'name' : tbl_test.name + "Test",
'promote' : True, 'caslib' : 'CASUSER({})'.format(authUser)})
try:
segment['session'].table.droptable(caslib='Public', name='lgdemoscore')
except:
pass
print("model Object", segment['modelObj'])
y_predicted = segment['session'].modelpublishing.runmodellocal(intable={'name': tbl_test.name + "Test",
"caslib" : 'CASUSER({})'.format(authUser)},
modelName= segment['modelObj'],
modeltable={'caslib': 'CASUSER({})'.format(authUser),'name': 'sas_model_table'},
outtable={'caslib':'Public','name':'lgdemoscore'})
segment['session'].table.droptable(caslib='CASUSER({})'.format(authUser),name='sas_model_table')
y_predicted = segment['session'].CASTable('lgdemoscore', caslib='Public').to_frame()
y_predicted.columns = y_predicted.columns.str.upper()
y_predicted = y_predicted['P_' + tab.children[0].children[3].children[0].children[0].value.upper() + '1'].values
segment['session'].table.droptable(caslib='CASUSER({})'.format(authUser),name=tbl_test.name + "Test")
fpr, tpr, _ = roc_curve(y_test, y_predicted)
ROCInfo = dict(FPR = fpr, Sensitivity = tpr)
segment['ROCInfo'] = ROCInfo
segment['LIFTInfo'] = create_lift(y_test, y_predicted)
accuracy = accuracy_score(y_test, [round(num) for num in y_predicted])
segment['misclassification'] = 1 - accuracy
else:
segment['ROCInfo'] = None
segment['LIFTInfo'] = None
segment['misclassification'] = None
yTestActual = y_test
yTestPredict = y_predicted
data = [(None, None),
(None, None),
(yTestActual, yTestPredict)]
# Define a directory where all the files for model manager will be kept
MODEL_DIR = 'automate/{}_{}-{}'.format(segment['model'], segment['region'], segment['handset'])
if os.path.exists(MODEL_DIR):
shutil.rmtree(MODEL_DIR)
os.makedirs(MODEL_DIR)
JSONFiles = pzmm.JSONFiles()
JSONFiles.calculateFitStat(data, MODEL_DIR)
JSONFiles.generateROCStat(data, target, MODEL_DIR)
JSONFiles.generateLiftStat(data, target, 1, MODEL_DIR)
segment['score_params'] = None
segment['assess_params'] = None