def by_sim_check(self, train=None, test=None, is_iter=False):
# exception handler
if train is None:
try:
train = self.train
except NameError:
self.split_data('random')
train = self.train
if test is None:
try:
test = self.test
except NameError:
self.split_data('random')
test = self.test
season_interval = self.param['season']
## checking sim
pred_res = []
ground_truth = []
for i in range(len(test)):
season_interval = self.param['season']
# season for test should be different (last season shouldn't be added)
if self.is_train:
test_season = [season_interval[0], season_interval[1]-1]
else:
test_season = [season_interval[0]+1, season_interval[1]]
test_list = dict_to_list(test[i]['data'],test_season)
ground_truth.append(dict_to_list(test[i]['data'],
[season_interval[1],
season_interval[1]])[0])
res = 0.0
errsum = 0.0
for j in range(len(train)):
train_list = dict_to_list(train[j]['data'], season_interval)
# this is only for current one. err should be array afterwards
if is_iter:
sii = SimCheckIteration(test_list,
train_list[0:-1], 'corr')
else:
sii = SimCheck(test_list, train_list[0:-1], 'corr')
err = sii.run()[0]
if err==0:
err = 0.00001
res+=(train_list[-1][0]/err)
errsum+=1/err
res/=errsum
pred_res.append(res)
eval_res = self.evaluation(pred_res, ground_truth)
return pred_res, eval_res
def by_regression(self, train=None, test=None):
# exception handler
if train is None:
try:
train = self.train
except NameError:
self.split_data('random')
train = self.train
if test is None:
try:
test = self.test
except NameError:
self.split_data('random')
test = self.test
season_interval = self.param['season']
# checking sim
input_season = [season_interval[0],season_interval[1]-1]
output_season = [season_interval[1],season_interval[1]]
train_input = []
train_output = []
for i in range(len(train)):
train_input.append([elm[0] for elm in dict_to_list(
train[i]['data'], input_season)])
train_output.append(dict_to_list(train[i]['data'],
output_season)[0])
rg_instance = Regressor(train_input,train_output)
coeff = rg_instance.run()
# season for test should be different (last season shouldn't be added)
test_input = []
test_output = []
for i in range(len(test)):
test_input.append([elm[0] for elm in dict_to_list(
test[i]['data'], input_season)])
test_output.append(dict_to_list(test[i]['data'],
output_season)[0])
predict_output = rg_instance.prediction(test_input, coeff[0])
eval_res = self.evaluation(predict_output, test_output)
return predict_output, eval_res
def by_sim_check(self, train=None, test=None, is_iter=False):
train, test = self.training_preparation(train, test)
season_interval = self.param['season']
# checking sim
train_input = train[list(set(train.columns) - set(self.ignore_fields))]
test_input = train[list(set(train.columns) - set(self.ignore_fields))]
pred_res = []
ground_truth = []
for i in range(len(test)):
# season for test should be different (last season shouldn't be added)
if self.is_train:
test_season = [season_interval[0], season_interval[1] - 1]
else:
test_season = [season_interval[0] + 1, season_interval[1]]
test_list = dict_to_list(test[i]['data'], test_season)
ground_truth.append(
dict_to_list(test[i]['data'],
[season_interval[1], season_interval[1]])[0])
res = 0.0
errsum = 0.0
for j in range(len(train)):
train_list = dict_to_list(train[j]['data'], season_interval)
# this is only for current one. err should be array afterwards
if is_iter:
sii = SimCheckIteration(test_list, train_list[0:-1],
'corr')
else:
sii = SimCheck(test_list, train_list[0:-1], 'corr')
err = sii.run()[0]
if err == 0:
err = 0.00001
res += (train_list[-1][0] / err)
errsum += 1 / err
res /= errsum
pred_res.append(res)
eval_res = self.evaluation(pred_res, ground_truth)
return pred_res, eval_res
