def test_cross_validation(self):
all_prediction_results = list()
for corresp_tissue in range(1, 24):
cv = CrossValidation(genes=self.genes,
all_gnids=self.gnids,
class_size=self.class_size,
fold_size=self.fold_size,
kmer_size=self.kmer_size,
exp_setting=self.exp_setting)
cv.build_datasets(assigned_genes=self.assigned_gnids,
neg_class_mode=self.exp_setting.get_neg_class_mode(),
corresp_tissue=corresp_tissue)
prediction_results = cv.validation()
all_prediction_results.append(prediction_results)
feature_vector = dict()
for gnid in self.gnids:
gnid_vector = list()
for prediction_results in all_prediction_results:
if prediction_results:
gnid_pr = prediction_results.get(gnid)
if gnid_pr is None:
gnid_vector.append('?')
else:
gnid_vector.append(gnid_pr.get_predicted_class())
feature_vector[gnid] = gnid_vector
def cross_validation(self):
# for each feature, build validation groups with fold size
print('Cross-Validation')
for feature in self.features:
print('Feature Name: {}, k-mer size: {}'.format(
feature.name, self.kmer_size))
# get wd_all_gnids per tissue
scid = feature.corresp_tissue
#wd_all_gnids_per_tissue = GetData.wd_all_gnid_per_tissue(self.exp_setting, scid)
cr_validation = CrossValidation(
genes=self.genes,
#all_gnids=wd_all_gnids_per_tissue,
all_gnids=None,
class_size=self.class_size,
fold_size=self.fold_size,
kmer_size=self.kmer_size,
exp_setting=self.exp_setting)
cr_validation.build_datasets(
assigned_genes=feature.assigned_genes,
neg_class_mode=self.exp_setting.get_neg_class_mode(),
corresp_tissue=feature.corresp_tissue)
# Do validation and get prediction results
prediction_results = cr_validation.validation()
# store prediction results in each feature
feature.set_prediction_results(
prediction_restuls=prediction_results)
# store confusion matrix in each feature
#feature.set_confusion_matrix_set(cm_set=cm_set)
feature.set_confusion_matrix_set(cm_set=self.set_confusion_matrix(
validation=cr_validation, fold_size=self.fold_size))
def start(self):
measurement = []
for i in range(self.repeat):
measure = []
splitted_dataset = cross.split_list(self.raw_data, self.k, True)
k_fold = cross.KFold(splitted_dataset)
for fold in k_fold:
stats, class_prob = NormalDist.calc_normal_stats(fold[0])
confusion_matrices = ConfusionMatrix.ConfusionMatrixStatistic(
stats)
for item in fold[1]:
confusion_matrices.add_result(
item[len(item) - 1],
NormalDist.get_class(stats, class_prob, item))
measure.append(confusion_matrices.calc_stats())
measurement.append(ConfusionMatrix.Measure.connect(measure))
return ConfusionMatrix.Measure.connect(measurement)
def start(self):
measurement = []
for i in range(self.repeat):
measure = []
splitted_dataset = cross.split_list(self.raw_data, self.k, True)
k_fold = cross.KFold(splitted_dataset)
for fold in k_fold:
stats, class_prob, buckets = edp.create_dictionary_with_buckets(
self.raw_data, fold[0], self.bins)
confusion_matrices = ConfusionMatrix.ConfusionMatrixStatistic(
stats)
for item in fold[1]:
confusion_matrices.add_result(
item[len(item) - 1],
edp.get_class(stats, class_prob, buckets, item))
measure.append(confusion_matrices.calc_stats())
measurement.append(ConfusionMatrix.Measure.connect(measure))
return ConfusionMatrix.Measure.connect(measurement)
def test_prediction_results(self):
all_prediction_results = list()
for corresp_tissue in range(1, 24):
cv = CrossValidation(genes=self.genes,
all_gnids=self.gnids,
class_size=self.class_size,
fold_size=self.fold_size,
kmer_size=self.kmer_size,
exp_setting=self.exp_setting)
cv.build_datasets(assigned_genes=self.assigned_gnids,
neg_class_mode=self.exp_setting.get_neg_class_mode(),
corresp_tissue=corresp_tissue)
prediction_results = cv.validation()
all_prediction_results.append(prediction_results)
feature_vector = dict()
for gnid in self.gnids:
gnid_vector = list()
for prediction_results in all_prediction_results:
if prediction_results:
gnid_pr = prediction_results.get(gnid)
if gnid_pr is None:
gnid_vector.append('?')
else:
gnid_vector.append(gnid_pr.get_predicted_class())
feature_vector[gnid] = gnid_vector
# show feature_vector
for tissue in range(1, 24):
print('Tissue#:', tissue)
for gnid, vector in feature_vector.items():
line = ",".join(str(value) for value in vector)
p_results = all_prediction_results[tissue - 1].get(gnid)
if p_results is None:
data_label = '?'
else:
data_label = 'data_label:{}'.format(p_results.get_assigned_class())
print("%s,%s,%s\n" % (gnid, line, data_label))
def test_features_dataset(self):
print('TEST features dataset')
for feature in self.features:
#print('Feature Name: {}'.format(feature.name))
#print('\tassigned genes: {}'.format(feature.assigned_genes))
# get wd_all_gnids per tissue
scid = feature.corresp_tissue
wd_all_gnids_per_tissue = GetData.wd_all_gnid_per_tissue(
self.exp_setting, scid)
cr_validation = CrossValidation(
genes=self.genes,
#all_gnids=self.wd_all_gnids,
all_gnids=wd_all_gnids_per_tissue,
class_size=self.class_size,
fold_size=self.fold_size,
kmer_size=self.kmer_size)
cr_validation.build_datasets(
assigned_genes=feature.assigned_genes,
neg_class_mode=self.exp_setting.get_neg_class_mode(),
corresp_tissue=feature.corresp_tissue)
cr_validation.test_datasets()