Question

我正试图通过对韩国电影的回顾来做出正面和负面的分类。数据已经过预处理。预处理数据由Google驱动器共享。通过参考示例LSTM模型进行配置。执行时，精度仅约为0.5。帮帮我......

此网址为train_test_dataset。 https://drive.google.com/file/d/1rnf84idV9pj72Ammu5VQ2_rNyGZIa3pU/view?usp=sharing 火车= 70000 测试=大约33000

from keras.models import Sequential
from keras.layers import Dense, Embedding,Dropout
from keras.layers import LSTM

max_features = 20000
maxlen = 300
batch_size = 32


import numpy as np
x_train = np.load('file_path')
y_train = np.load('file_path')
x_test = np.load('file_path')
y_test = np.load('file_path')


print('train_x shape:', train_x.shape)
print('test_x shape: ', test_x.shape)



print('model generate...')
model = Sequential()
model.add(Embedding(max_features, 128))
model.add(LSTM(128, dropout=0.2, recurrent_dropout=0.2))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy',
              optimizer='adam',
              metrics=['accuracy'])



try:
    print('Train...')
    model.fit(train_x, train_y,
              batch_size=batch_size,
              epochs=1,
              validation_data=(test_x, test_y))
    score, acc = model.evaluate(test_x, test_y,
                                batch_size=batch_size)
    print('Test score:', score)
    print('Test accuracy:', acc)
except Exception as e:
    print(e)

我的LSTM模型精度仅为0.5左右

0 个答案: