def analyze_maj(fname, mode):
input_dir = 'C:\\research\\falseMedicalClaims\\White and Hassan\\model input\\majority\\'
reports_dir = input_dir + 'reports\\'
reports_filename = fname + '_' + mode.name + '_report'
gen_majority_report(input_dir + fname + '.csv',
reports_dir + reports_filename + '.csv', mode)
df = pd.read_csv(reports_dir + '\\' + reports_filename + '.csv')
queries = get_queries_from_df(df)
labels = {q: int(queries[q].label) for q in queries}
gen_metrics_comparison(folder=reports_dir,
query_filenames=[reports_filename],
actual_values=labels,
cmp_filename=reports_filename + '_stats_report',
mode=mode)
def group_all():
#input_dir = 'C:\\research\\falseMedicalClaims\\ECAI\\model input\\Yael_sigal_Irit\\by_group'
#feature_file = "rel_only_group_features_by_stance_citation_range_1"
#feature_file = "group_features_by_stance_citation_range_1"
#df = pd.read_csv(input_dir + '\\group_features_by_stance.csv')
#df = pd.read_csv(input_dir + '\\group_features_by_stance_no_enum.csv')
#feature_file = "group_features_by_stance_citation_range_only_clinical1"
#feature_file = "group_features_by_stance_citation_range_only_rev1"
#feature_file = "group_features_by_stance_citation_range_1_no_stance"
#feature_file = "group_features_by_stance_citation_range_1_no_stance_no_rel"
input_folder = 'C:\\research\\falseMedicalClaims\\ECAI\\model input\\'
cls = 'all_equal_weights'
input_dir = input_folder + cls + '\\by_group'
feature_file = "group_features_by_stance"
df = pd.read_csv(input_dir + '\\' + feature_file + '.csv')
queries = get_queries_from_df(df)
labels = {q: int(queries[q].label) for q in queries}
#mc = MajorityClassifier(input_dir + '\\majority.csv')
decisionTreeLearner1 = SKLearner(DecisionTreeClassifier(random_state=0))
svcLearner = SKLearner(svm.SVC(gamma='scale'))
#layers = [Layer(input=5, output=10), Layer(input=10, output=10)]
#layers = [Layer(input=31, output=20), Layer(input=20, output=10)]
layers = [Layer(input=39, output=20), Layer(input=20, output=10)]
net = TwoLayersNet(layers)
#params = get_parms(net)
#nnlearner = NNLearner(dataHelper.Method.GROUP_ALL, net=net, params=params)
learners = [decisionTreeLearner1]
#learners = [nnlearner]
predictions = test_models(learners, queries, dataHelper.Split.BY_QUERY,
dataHelper.Method.GROUP_ALL)
reports_dir = input_dir + '\\reports\\'
query_report_file_name = feature_file + '_query_report.csv'
query_report_full_file_name = reports_dir + query_report_file_name
#metrics_report_file_name =input_dir + '\\reports\\'+feature_file+'_metrics_report.csv'
create_query_report_file(query_report_full_file_name, queries, learners,
predictions, labels)
files = ['google labels', 'majority', query_report_file_name]
label_file = input_folder + cls + '\\labels.csv'
gen_all_metrics_comparison(folder=reports_dir,
files=files,
label_file=label_file)
def w_h_report():
input_dir = 'C:\\research\\falseMedicalClaims\\White and Hassan\\model input\\mult_features\\'
feature_file = "h_index_stance_label_shrink_neg"
feature_file = "h_index_stance_label_shrink_neg"
df = pd.read_csv(input_dir + '\\' + feature_file + '.csv')
# feature_file = "group_features_by_stance_shrink"
reports_dir = input_dir + '\\reports\\'
query_report_file_name = feature_file + '_query_report.csv'
queries = get_queries_from_df(df)
labels = {q: int(queries[q].label) for q in queries}
files = ['majority_nol', query_report_file_name]
gen_metrics_comparison(folder=reports_dir,
query_filenames=files,
actual_values=labels,
cmp_filename=feature_file + 'stats_report',
mode=ValToClassMode.THREE_CLASSES_PESSIMISTIC)
gen_metrics_comparison(folder=reports_dir,
query_filenames=files,
actual_values=labels,
cmp_filename=feature_file + 'stats_report',
mode=ValToClassMode.THREE_CLASSES_OPTIMISTIC)
def w_h():
# input_folder = 'C:\\research\\falseMedicalClaims\\White and Hassan\\model input\\'
# input_dir = 'C:\\research\\falseMedicalClaims\\White and Hassan\\model input\\mult_features\\no outlayers\\'
# feature_file = "group_features_by_stance_shrink_nol"
input_dir = 'C:\\research\\falseMedicalClaims\\White and Hassan\\model input\\mult_features\\'
feature_file = "group_features_by_stance_nol"
# input_dir = 'C:\\research\\falseMedicalClaims\\White and Hassan\\model input\\posterior\\'
# feature_file = "weighted_posterior_normed_ratio"
df = pd.read_csv(input_dir + '\\' + feature_file + '.csv')
features = list(df.head())[2:]
queries = get_queries_from_df(df)
labels = {q: int(queries[q].label) for q in queries}
decisionTreeLearner = SKLearner(DecisionTreeClassifier(random_state=0),
features)
gnb = SKLearner(GaussianNB())
ecl = ExpectedValLearner()
mult = MultipleBinaryCls(ValToClassMode.THREE_CLASSES_PESSIMISTIC)
neigh = SKLearner(KNeighborsClassifier(n_neighbors=5))
lr = SKLearner(LogisticRegression(C=1e5))
kmeans = KMeansClassifier(input_dir + '\\dist.csv')
decisionforestLearner = SKLearner(RandomForestClassifier(random_state=0))
#learners = [kmeans, neigh]
learners = [mult]
predictions = test_models(learners, queries, dataHelper.Split.BY_QUERY,
dataHelper.Method.GROUP_ALL,
ValToClassMode.THREE_CLASSES_PESSIMISTIC)
reports_dir = input_dir + '\\reports\\'
query_report_file_name = feature_file + '_query_report.csv'
query_report_full_file_name = reports_dir + query_report_file_name
create_query_report_file(query_report_full_file_name, queries, learners,
predictions, labels, ValToClassMode.W_H)
files = ['majority', query_report_file_name]
gen_metrics_comparison(folder=reports_dir,
query_filenames=files,
actual_values=labels,
cmp_filename=feature_file + 'stats_report',
mode=ValToClassMode.W_H)
def group_all():
#input_dir = 'C:\\research\\falseMedicalClaims\\examples\\model input\\pubmed\\normed\\group7'
input_dir = 'C:\\research\\falseMedicalClaims\\ECAI\\model input\\Yael\\by_group'
models = [ #DecisionTreeRegressor(random_state=0),
DecisionTreeClassifier(random_state=0)
]
#LinearRegression()]
#df = pd.read_csv(input_dir + '\\group_features.csv')
df = pd.read_csv(input_dir + '\\group_features_by_stance.csv')
#df = pd.read_csv(input_dir + '\\group_features_by_paper_type.csv')
queries = get_queries_from_df(df)
labels = {q: int(queries[q].label) for q in queries}
mc = MajorityClassifier(input_dir + '\\majority.csv')
predictions = test_models(models, queries, dataHelper.Split.BY_QUERY,
dataHelper.Method.GROUP_ALL)
create_report_file(input_dir + '\\group_features_by_stance_report.csv',
queries=queries,
models=models,
predictions=predictions,
majority_classifier=mc,
labels=labels)
def ijcai():
#input_folder = 'C:\\research\\falseMedicalClaims\\IJCAI\\model input\\ecai_new'
input_folder = 'C:\\research\\falseMedicalClaims\\IJCAI\\model input\\non'
#input_folder = 'C:\\research\\falseMedicalClaims\\IJCAI\\model input\\GTIC'
input_dir = input_folder + '\\by_group'
feature_file = "dummy_added_group_features_by_stance3"
#feature_file = "dummy_added_group_features_by_stance_exp"
#feature_file = "dist_exp"
#feature_file = "dist_group_features_by_stance_paste_ecai2"
feature_file = "dist"
df = pd.read_csv(input_dir + '\\' + feature_file + '.csv')
features = list(df.head())[2:]
queries = get_queries_from_df(df)
labels = {q: int(queries[q].label) for q in queries}
decisionTreeLearner = SKLearner(DecisionTreeClassifier(random_state=0),
features)
gnb = SKLearner(GaussianNB())
ecl = ExpectedValLearner()
mult = MultipleBinaryCls(ValToClassMode.W_H)
neigh = SKLearner(KNeighborsClassifier(n_neighbors=5))
kmeans = KMeansClassifier(input_dir + '\\dist.csv')
#decisionforestLearner = SKLearner(RandomForestClassifier(random_state=0))
learners = [mult]
#learners = [decisionforestLearner]
predictions = test_models(learners, queries, dataHelper.Split.BY_QUERY,
dataHelper.Method.GROUP_ALL)
reports_dir = input_dir + '\\reports\\'
query_report_file_name = feature_file + '_query_report.csv'
query_report_full_file_name = reports_dir + query_report_file_name
create_query_report_file(query_report_full_file_name, queries, learners,
predictions, labels)
files = ['majority', query_report_file_name]
gen_all_metrics_comparison(folder=reports_dir,
files=files,
actual_values=labels,
cmp_filename=feature_file + 'stats_report')
def learn_by_doctors_annotations(val2class,
feature_file,
directory,
resample,
majority_filename,
quick=False,
filter_queries=None):
input_dir = 'C:\\research\\falseMedicalClaims\\White and Hassan\\model input\\mult_features\\'
output_dir = 'C:\\research\\falseMedicalClaims\\White and Hassan\\model input\\' + directory + '\\'
# feature_file = "group_features_by_stance_nol"
#majority_file = input_dir+ majority_filename
reports_dir = output_dir + '\\reports\\'
query_report_file_name = feature_file + '_query_report.csv'
query_report_full_file_name = reports_dir + query_report_file_name
#majority_report_file_name = 'majority_' + val2class.name + '.csv'
#majority_report_full_file_name = reports_dir + majority_report_file_name
#gen_majority_report(majority_file, majority_report_full_file_name, val2class)
#feature_file = "group_features_by_label_shrink_nol"
#feature_file = "group_features_by_stance_label_shrink_nol"
df = pd.read_csv(input_dir + '\\' + feature_file + '.csv')
queries = get_queries_from_df(df)
labels = {q: int(queries[q].label) for q in queries}
knn = KNeighborsClassifier(n_neighbors=5)
neigh = SKLearner(knn, features=None, resample=resample)
#weights = {1: 4.0, 3: 1.5 , 5:1.0}
weights = {1: 100, 3: 25, 5: 1}
svc = SKLearner(svm.SVC(class_weight=weights))
rfc = RandomForestClassifier(random_state=0)
decisionforestLearner = SKLearner(rfc, features=None, resample=resample)
# lr = SKLearner(LinearRegression(C=1e5))
mult = MultipleBinaryCls(ValToClassMode.FOUR_CLASSES, resample=resample)
#learners = [decisionforestLearner, neigh]
maj = MajorityClassifier(val2class)
learners = [
maj,
MultipleCls(val2class, resample=resample), decisionforestLearner, neigh
]
# learners = [mult, decisionforestLearner, MultipleCls(val2class, resample = resample) ]
if quick:
for learner in learners:
learner.quick_learn(queries)
predictions = test_models(learners, queries, dataHelper.Split.BY_QUERY,
dataHelper.Method.GROUP_ALL, val2class)
create_query_report_file(query_report_full_file_name, input_dir,
feature_file, queries, learners, predictions,
labels, val2class)
# files = [majority_report_file_name, query_report_file_name]
files = [query_report_file_name]
if resample:
cmp_filename = feature_file + '_stats_report_resample'
else:
cmp_filename = feature_file + '_stats_report'
gen_all_metrics_comparison(folder=reports_dir,
files=files,
actual_values=labels,
cmp_filename=cmp_filename,
val2class=val2class)
if filter_queries:
gen_all_metrics_comparison(folder=reports_dir,
files=files,
actual_values=labels,
cmp_filename=cmp_filename,
val2class=val2class,
filter_queries=filter_queries)