def main():
feature = Feature()
tests = []
with open('./data/basis.txt') as p:
for test_line in p:
tests.append(feature.convert_sentence(test_line))
max_width = 35 # TODO: 自動で取れるように
tests = [(test + [0] * (max_width - len(test)))[:max_width]
for test in tests]
svc = joblib.load('./model/svc.pkl')
prediction = svc.predict(tests)
with open('./data/output', mode='w') as p:
p.write("\n".join([
'1' if result == '+1' else '-1' for result in prediction.tolist()
]))
def main():
feature = Feature()
trains, labels = [], []
with open('data/train.list') as p:
for train_line in p:
label, _, train_txt = train_line.split(maxsplit=2)
trains.append(feature.convert_sentence(train_txt))
labels.append(label)
max_width = int(statistics.mean([len(train) for train in trains]) * 2)
trains = [(train + [0] * (max_width - len(train)))[:max_width] for train in trains]
assert statistics.mean([len(train) for train in trains]) == max_width, '配列の要素数が次元ごとで揃っていない'
assert len(trains) == len(labels), '学習データとラベルの数が不一致'
numpy.savez('data/dataset.npz',
trains=numpy.array(trains, dtype=numpy.float),
labels=numpy.array(labels, dtype=numpy.str))