def ECC_test_2_fold(data, label, random_state=3071980, ensemble=5):
# data set information
n_label = label.shape[1]
# split training and test data set
X_train, y_train, X_test, y_test = iterative_train_test_split(
np.matrix(data), np.matrix(label), test_size=0.5)
X_train = pd.DataFrame(X_train, columns=data.columns)
X_test = pd.DataFrame(X_test, columns=data.columns)
y_train = pd.DataFrame(y_train, columns=label.columns)
y_test = pd.DataFrame(y_test, columns=label.columns)
performance_df_all = pd.DataFrame()
for j in range(2):
X_test, X_train = X_train, X_test
y_test, y_train = y_train, y_test
# ensemble
y_pred_ensemble = pd.DataFrame(np.zeros(y_test.shape),
columns=y_test.columns,
index=y_test.index)
y_prob_ensemble = pd.DataFrame(np.zeros(y_test.shape),
columns=y_test.columns,
index=y_test.index)
for i in range(ensemble):
# training
# print("--- start training ---\n")
classifier_list, training_time, order = naiveBayes_multi_label_training(
X_train, y_train)
# testing
# print("--- start testing ---\n")
y_predict, y_prob, testing_time = naiveBayes_multi_label_testing(
X_test, n_label, classifier_list, order)
y_predict.columns = label.columns[order]
y_prob.columns = label.columns[order]
y_predict = y_predict[label.columns]
y_prob = y_prob[label.columns]
y_pred_ensemble = y_pred_ensemble + y_predict
y_prob_ensemble = y_prob_ensemble + y_prob
y_pred_ensemble = (((y_pred_ensemble / ensemble) >= 0.5) *
1).astype('int')
y_prob_ensemble = y_prob_ensemble / ensemble
# evaluation
performance = evaluation(y_pred_ensemble, y_prob_ensemble, y_test)
performance_df = pd.DataFrame.from_dict(performance, orient='index')
#performance_df_all.index = performance_df.index
performance_df_all = pd.concat([performance_df_all, performance_df],
axis=1)
return performance_df_all
def BCC_test_structure_twofold(data,
label,
dataPath,
random_state=3071980,
ensemble=5,
structure="random"):
# data set information
n_label = label.shape[1]
n_attr = data.shape[1]
n_instance = data.shape[0]
# get order
if structure == "DAG":
model = BayesianNetwork.from_samples(label, algorithm='greedy')
bayes_net = get_structure(model, label.columns)
order = get_order(model, label.columns)
# split training and test data set
X_train, y_train, X_test, y_test = iterative_train_test_split(
np.matrix(data), np.matrix(label), test_size=0.5)
X_train = pd.DataFrame(X_train, columns=data.columns)
X_test = pd.DataFrame(X_test, columns=data.columns)
y_train = pd.DataFrame(y_train, columns=label.columns)
y_test = pd.DataFrame(y_test, columns=label.columns)
performance_df_all = pd.DataFrame()
for j in range(2):
X_test, X_train = X_train, X_test
y_test, y_train = y_train, y_test
# ensemble
y_pred_ensemble = pd.DataFrame(np.zeros(y_test.shape),
columns=y_test.columns,
index=y_test.index)
y_prob_ensemble = pd.DataFrame(np.zeros(y_test.shape),
columns=y_test.columns,
index=y_test.index)
for i in range(ensemble):
if structure == "tree":
if len(label.columns) <= ensemble:
root_index = i
else:
root_index = random.randint(0, len(label.columns) - 1)
model = BayesianNetwork.from_samples(label,
algorithm='chow-liu',
root=root_index)
bayes_net = get_structure(model, label.columns)
order = get_order(model, label.columns)
# training
# print("--- start training ---\n")
classifier_list, learned_label = naiveBayes_multi_label_training_order(
X_train, y_train, bayes_net, order)
# testing
# print("--- start testing ---\n")
y_predict, y_prob = naiveBayes_multi_label_testing_order(
X_test, n_label, classifier_list, bayes_net, learned_label)
y_predict = y_predict[label.columns]
y_prob = y_prob[label.columns]
y_pred_ensemble = y_pred_ensemble + y_predict
y_prob_ensemble = y_prob_ensemble + y_prob
y_pred_ensemble = (((y_pred_ensemble / ensemble) >= 0.5) *
1).astype('int')
y_prob_ensemble = y_prob_ensemble / ensemble
y_pred_ensemble = y_pred_ensemble.fillna(0)
y_prob_ensemble = y_prob_ensemble.fillna(0)
# evaluation
performance = evaluation(y_pred_ensemble, y_prob_ensemble, y_test)
performance_df = pd.DataFrame.from_dict(performance, orient='index')
performance_df_all = pd.concat([performance_df_all, performance_df],
axis=1)
return performance_df_all
def BCC_test_2_fold(data,
label,
dataPath,
bayes_net,
random_state=3071980,
ensemble=5,
root=None):
# data set information
n_label = label.shape[1]
# split training and test data set
X_train, y_train, X_test, y_test = iterative_train_test_split(
np.matrix(data), np.matrix(label), test_size=0.5)
X_train = pd.DataFrame(X_train, columns=data.columns)
X_test = pd.DataFrame(X_test, columns=data.columns)
y_train = pd.DataFrame(y_train, columns=label.columns)
y_test = pd.DataFrame(y_test, columns=label.columns)
performance_df_all = pd.DataFrame()
for j in range(2):
X_test, X_train = X_train, X_test
y_test, y_train = y_train, y_test
# ensemble
y_pred_ensemble = pd.DataFrame(np.zeros(y_test.shape),
columns=y_test.columns,
index=y_test.index)
y_prob_ensemble = pd.DataFrame(np.zeros(y_test.shape),
columns=y_test.columns,
index=y_test.index)
node_list = []
for node, par in bayes_net.items():
if par != set():
node_list.append(node)
en = 0
for i in range(ensemble):
if root != None:
root_name = root
else:
root_name = label.columns[random.randint(
0, label.shape[1] - 1)]
if root_name not in node_list:
continue
else:
# training
# print("--- start training ---\n")
classifier_list, learned_label = naiveBayes_multi_label_training(
X_train, y_train, bayes_net, root_name)
# testing
# print("--- start testing ---\n")
y_predict, y_prob = naiveBayes_multi_label_testing(
X_test, n_label, classifier_list, bayes_net,
learned_label)
y_predict = y_predict[label.columns]
y_prob = y_prob[label.columns]
y_pred_ensemble = y_pred_ensemble + y_predict
y_prob_ensemble = y_prob_ensemble + y_prob
en += 1
y_pred_ensemble = (((y_pred_ensemble / en) >= 0.5) * 1).astype('int')
y_prob_ensemble = y_prob_ensemble / en
y_pred_ensemble = y_pred_ensemble.fillna(0)
y_prob_ensemble = y_prob_ensemble.fillna(0)
# evaluation
performance = evaluation(y_pred_ensemble, y_prob_ensemble, y_test)
performance_df = pd.DataFrame.from_dict(performance, orient='index')
performance_df_all = pd.concat([performance_df_all, performance_df],
axis=1)
return performance_df_all
def BCC_test_order_twofold(data, label, dataPath, bayes_net, random_state=3071980, ensemble=5, order_method="random"):
# data set information
n_label = label.shape[1]
n_attr = data.shape[1]
n_instance = data.shape[0]
avg_label_per_instance = label.sum(axis=1).mean()
# get order
if order_method == "best_prediction":
y_predict, y_test = BR_test(data, label, dataPath, 3071980)
acc = (y_predict.values == y_test.values).mean(axis=0)
order = list(label.columns[np.argsort(-acc)])
elif order_method == "largest_edges":
a = [(x, len(y)) for x, y in bayes_net.items()]
a_sort = sorted(a, key=lambda x: x[1], reverse=True)
order = [x[0] for x in a_sort]
# split training and test data set
X_train, y_train, X_test, y_test = iterative_train_test_split(np.matrix(data), np.matrix(label), test_size=0.5)
X_train = pd.DataFrame(X_train, columns=data.columns)
X_test = pd.DataFrame(X_test, columns=data.columns)
y_train = pd.DataFrame(y_train, columns=label.columns)
y_test = pd.DataFrame(y_test, columns=label.columns)
performance_df_all = pd.DataFrame()
for j in range(2):
X_test, X_train = X_train, X_test
y_test, y_train = y_train, y_test
# ensemble
y_pred_ensemble = pd.DataFrame(np.zeros(y_test.shape), columns=y_test.columns, index=y_test.index)
y_prob_ensemble = pd.DataFrame(np.zeros(y_test.shape), columns=y_test.columns, index=y_test.index)
for i in range(ensemble):
if order_method == "random":
order = random.sample(list(range(n_label)), n_label) # get orders
# training
# print("--- start training ---\n")
classifier_list, learned_label = naiveBayes_multi_label_training_order(X_train, y_train, bayes_net, order)
# testing
# print("--- start testing ---\n")
y_predict, y_prob = naiveBayes_multi_label_testing_order(X_test, n_label, classifier_list, bayes_net,
learned_label)
y_predict = y_predict[label.columns]
y_prob = y_prob[label.columns]
y_pred_ensemble = y_pred_ensemble + y_predict
y_prob_ensemble = y_prob_ensemble + y_prob
y_pred_ensemble = (((y_pred_ensemble / ensemble) >= 0.5) * 1).astype('int')
y_prob_ensemble = y_prob_ensemble / ensemble
y_pred_ensemble = y_pred_ensemble.fillna(0)
y_prob_ensemble = y_prob_ensemble.fillna(0)
# evaluation
performance = evaluation(y_pred_ensemble, y_prob_ensemble, y_test)
performance_df = pd.DataFrame.from_dict(performance, orient='index')
performance_df_all = pd.concat([performance_df_all, performance_df],axis=1)
return performance_df_all