def __init__(self):
self.caseid = 0
self.bayes_score = 50 # init value of bayes score
self.numRadarElem = 6 # number of elements in radar plot
self.radar = [50] * self.numRadarElem # init value of radar plot
self.radarName = [
'Attention', 'Self Control', 'Speed', 'Sensitivity',
'Cautiousness', 'Steadiness'
]
self.percentileElem = 3 # number of elements in percentile plot
self.percentile = [
50
] * self.percentileElem # init value of percentile plot
self.percentileName = [
'Concentration', 'Stability', 'Focus Continuity'
]
self.motion_box_numpoints = 3900 # not working at a fixed length now
self.motion_box = np.zeros((self.motion_box_numpoints, 2))
self.motion_index_points = 390 # not working at a fixed length now
self.motion_index = [0] * self.motion_index_points
self.roseBinNum = 360
self.roseValue = [0] * self.roseBinNum
self.cpt_response_min = [
np.arange(52) * 1.5 * 10 / 60, [], []
] # list0 for correct, list1 for commmission, list2 for omission
self.db = db_manager()
def main(testCaseIds, testdb_name='webdarintest', write_db=False):
""" Calculate features for incoming test cases.
Args:
testCaseIds: list. Raw data for case ids in the array must be found in testdb.
testdb_name: name of the testing database to fetch raw data and store features. Should contain raw data (hmd_data) for test cases.
write_db: if true, write result to database (features for testCaseIds)
"""
# obtain raw data
mydb = db_manager(testdb_name)
data = get_hmd_data(testCaseIds, mydb)
# compute features
features = process_feature(testCaseIds, data)
# insert to db
if write_db:
insert_features(features, mydb)
return features
def get_individual_data(CaseIds,modelIOStruct = None, tableNames=None,fieldNames=None,whereClauses=None, primary_keys = None):
isTrainingSet = CaseIds[:,1]
numTest = np.sum(isTrainingSet==0)
CaseIds = CaseIds[:,0]
try:
if modelIOStruct != None:
tableNames,fieldNames,whereClauses, primary_keys = modelIOStruct.getInput()
if (tableNames == None) and (fieldNames == None):
# prepare a cursor object using cursor() method
dataCPT = get_cpt(CaseIds, isTrainingSet, cursor_rnd, cursor_web)
dataSignal = get_signal(CaseIds, isTrainingSet, cursor_rnd, cursor_web)
# dataSNAP = get_snap(CaseIds, db, cursor)
dataSNAP = np.zeros((len(CaseIds),1))
# dataADHD = get_adhd([CaseIds[i] for i in np.where(isTrainingSet==1)[0]], db_rnd, cursor_rnd, cross_db = True)
# testADHD = np.concatenate((np.asarray([CaseIds[i] for i in np.where(isTrainingSet==0)[0]]).reshape(numTest,1),np.zeros((numTest,1))),axis=1)
# dataADHD = np.concatenate((dataADHD,testADHD),axis=0)
dataADHD = get_adhd(CaseIds, isTrainingSet, cursor_rnd, cursor_web, cross_db=True)
result = np.concatenate((dataADHD, dataCPT, dataSignal, dataSNAP), axis = 1)
elif len(tableNames) != len(fieldNames):
print('tableNames and fieldNames must be the same length')
raise Exception('tableNames and fieldNames must be the same length')
else:
mydb = db_manager()
results = []
for dbname,isTrain in zip((traindb_name,testdb_name),[1,0]):
mydb.connect(dbname)
df = get_model_input(modelIOStruct,CaseIds[isTrainingSet==isTrain],mydb)
results.append(df)
result = pd.concat(results,axis=0)
except Exception as e:
raise Exception(4)
return result
def insert_bayes_probabilities_df(df, caseIds): """ Insert probabilities dataframe to database, for caseIds. """ mydb = db_manager(testdb_name) mydb.insert_table(df,'bayes_probabilities',index_name='CaseId',del_row_if_exist = True)
def main(testCaseIds,
traindb_name='vrclassroomdata',
testdb_name='webtest',
use_training_cases=False,
write_db=False,
train_features=None,
test_features=None):
""" calculate percentile for incoming test cases, based on training set
Args:
testCaseIds: array-like. Features for case ids in the array must be found in testdb.
traindb_name: name of the training database to fetch training set. Should contain features for training cases
testdb_name: name of the testing database to store percentile. Should contain features for test cases.
use_training_cases: if true, training features are the used for testing as well (append to the end). ``traindb_name`` and ``testdb_name`` must be the same for this to take effect.
write_db: if true, write result to database (percentile for testCaseIds)
train_features: if specified, use the input data instead of fetching from traindb_name
test_features: if specified, use the input data instead of fetching from testdb_name
"""
# input check
if use_training_cases and (traindb_name != testdb_name):
raise ValueError(
"``traindb_name`` and ``testdb_name`` must be the same")
# initiate db connection
mydb_train = db_manager(traindb_name)
mydb_test = db_manager(testdb_name)
# define field names for input and output
input_tables = ['head_features', 'cpt_output_results']
input_fields = [['PathLen', 'TimeActive', 'NumRot', 'TotalDeg'],
[
'OmissionErrors', 'CommissionErrors', 'TargetsRT',
'TargetsRtVariability', 'CommissionErrorsRT',
'CommissionErrorsRtVariability'
]]
input_where = ['', 'where block=0']
input_primarykey = ['CaseId', 'CaseId']
output_fields = []
for input_field in input_fields:
output_fields += ['Per' + x for x in input_field]
output_fields += ['CaseId', 'BlockNum']
# get training caseids
trainCaseIds = get_training_ids(mydb_train)
if len(trainCaseIds) == 0:
raise RuntimeError('cannot get training ids from training set table')
if use_training_cases:
testCaseIds += trainCaseIds
# fetch features for training & testing
modIO = modelIO(input_tables, input_fields, input_where, input_primarykey)
if train_features is None:
train_features = get_model_input(modIO, trainCaseIds, mydb_train)
if test_features is None:
test_features = get_model_input(modIO, testCaseIds, mydb_test)
# calculate percentile for all fields
percentiles = []
for field in modIO.getAllFields():
percentiles.append(
percentileCalc(train_features[field], test_features[field]))
percentiles.append(testCaseIds)
percentiles.append(np.zeros(len(testCaseIds)))
percentiles = np.stack(percentiles, axis=1)
result = pd.DataFrame(percentiles,
columns=output_fields,
index=testCaseIds)
# insert to db
if write_db:
insert_percentile(result, mydb_test)
return result