def fit(ml_algorithm, train_files, train_labels, name):
train_features, _ = FeatureExtractor.get_features(train_files, [])
try:
algorithm_type = ALGORITHMS[ml_algorithm]
except KeyError:
print 'Algorithm type not valid!'
return
mla = algorithm_type()
mla.fit(train_features, train_labels, name)
def get_probs(ml_algorithm, train_files, test_files, name):
_, test_features = FeatureExtractor.get_features(train_files, test_files)
try:
algorithm_type = ALGORITHMS[ml_algorithm]
except KeyError:
print 'Algorithm type not valid!'
return
mla = algorithm_type()
prob_ind, highest_score = mla.predict(test_features, name)
return prob_ind, highest_score
def train_and_predict(ml_algorithm, train_files, train_labels, test_files, test_labels):
train_features, test_features = FeatureExtractor.get_features(train_files, test_files)
try:
algorithm_type = ALGORITHMS[ml_algorithm]
except KeyError:
print 'Algorithm type not valid!'
return
mla = algorithm_type()
score, indices, probs = mla.fit_and_predict(train_features, test_features, train_labels, test_labels)
print 'Score: {}'.format(score)
return indices, probs
def train_kfold(ml_algorithm, source_code, labels, ast_nodes=None):
source_code = np.array(source_code)
labels = np.array(labels)
#ast_nodes = np.array(ast_nodes)
try:
algorithm_type = ALGORITHMS[ml_algorithm]
except KeyError:
print 'Algorithm type not valid!'
return
start = time.time()
n_trees = [400]
for nt in n_trees:
mla = algorithm_type(n_trees=nt)
print 'nt: {}'.format(nt)
k = 10
code_per_author = 10
accuracy = 0
for i in range(k):
test_indices = []
it = code_per_author / k
for j in range(it):
test_indices.extend(np.array(range(i*it+j, len(source_code), code_per_author)))
train_indices = []
for sci in range(len(source_code)):
if sci not in test_indices:
train_indices.append(sci)
train_features, test_features = FeatureExtractor.get_features(source_code[train_indices],
source_code[test_indices])
score = mla.fit_and_predict(train_features, test_features, labels[train_indices], labels[test_indices])
print 'Score after {}th fold is {}'.format(i, score)
accuracy += score
print 'Final score: {}'.format(accuracy / float(k))
end = time.time() - start
print 'Execution time: {}'.format(end)
print '-----------------------------------------------'
