from sklearn.neighbors import KNeighborsClassifier
from utils import data_utils, vectorize
class KNN:
def __init__(self, vectorizer, data_fname='../../data/imdb.csv', n_neighbors=10):
self.data_fname = data_fname
self.n_neighbors = n_neighbors
self.vectorizer = vectorizer
def run(self):
X_train, X_test, y_train, y_test = data_utils.load_train_test_data(self.data_fname)
train_features, test_features = self.vectorizer.feature_extraction(X_train, X_test)
neigh = KNeighborsClassifier(n_neighbors=self.n_neighbors)
neigh.fit(train_features, y_train)
print(neigh.score(test_features, y_test))
if __name__ == '__main__':
vectorizer = vectorize.Vectorizer('BOW', ngram_range=(1, 1))
model = KNN(vectorizer=vectorizer)
model.run()
if test_acc / X_test.shape[0] > max_acc:
max_acc = test_acc / X_test.shape[0]
step = i
else:
stop_num += 1
if stop_num == self.para['stop_num']:
break
print('Best Performance: %.3f at Epoch %d' % (max_acc, step))
if __name__ == '__main__':
vectorizer = vectorize.Vectorizer(
'Word2Vec',
emb_fname='/data/linpq/Word2Vec/glove.840B.300d.txt',
word_index_fname='../../data/imdb_word_index.json')
para = {
'learning_rate': 0.001,
'l2_reg': 1e-5,
'hidden_size': 300,
'max_seq_len': 500,
'num_layers': 1,
'epoch_num': 100,
'batch_size': 200,
'stop_num': 10
}
classifier = Classifier(vectorizer=vectorizer, para=para)
classifier.run()
test_acc / X_test.shape[0]))
if test_acc / X_test.shape[0] > max_acc:
max_acc = test_acc / X_test.shape[0]
step = i
else:
stop_num += 1
if stop_num == self.para['stop_num']:
break
print('Best Performance: %.3f at Epoch %d' % (max_acc, step))
if __name__ == '__main__':
vectorizer = vectorize.Vectorizer(
'Word2Vec',
emb_fname='../../data/imdb.vec',
word_index_fname='../../data/imdb_word_index.json')
para = {
'learning_rate': 0.001,
'l2_reg': 1e-4,
'hidden_size': 300,
'max_seq_len': 500,
'epoch_num': 100,
'batch_size': 200,
'stop_num': 10
}
classifier = Classifier(vectorizer=vectorizer, para=para)
classifier.run()