【NLP】文本分類算法-基于字符級(jí)的無(wú)詞嵌入雙向循環(huán)神經(jīng)網(wǎng)絡(luò)（雙向 GRU）

import os import pandas as pdpath  = '/Users/wuzhengxiang/Documents/DataSets/TextCnn'os.chdir(path)data = pd.read_csv('text_all.csv')
#對(duì)數(shù)據(jù)進(jìn)行隨機(jī)打亂data = data.sample(frac=1, random_state=42)print(data.shape)(19670, 2)
#查看0-1的比例,可以看出來(lái)，數(shù)據(jù)集基本上平衡data['label'].value_counts()1    98820    9788
#查看前10行的數(shù)據(jù)data.head(10)                         text  label17036        鄭 29526 Q 77544      15426                    葩葩葩l      014173            網(wǎng)站盤(pán)需要買的私聊我.      114582               買家秀和賣家秀？01730                  1776看v      01444              我又沒(méi)送你謝我干啥?      010439            7645 55562筘      02448   伽韋 sx111505 珂視頻箹   Ku      110423            影薇 w2753636      111782      胸還沒(méi)有寒磊的 大?  還奶子疼！0

from keras.preprocessing.text import Tokenizerfrom keras.preprocessing.text import text_to_word_sequencefrom keras.preprocessing.sequence import pad_sequences
from keras.models import Model from keras.models import Sequential
from keras.layers import Input, Dense, Embedding, Conv1D, Conv2D, MaxPooling1D, MaxPool2Dfrom keras.layers import Reshape, Flatten, Dropout, Concatenatefrom keras.layers import SpatialDropout1D, concatenatefrom keras.layers import GRU, Bidirectional, GlobalAveragePooling1D, GlobalMaxPooling1D
from keras.callbacks import Callbackfrom keras.optimizers import Adam
from keras.callbacks import ModelCheckpoint, EarlyStoppingfrom keras.models import load_modelfrom keras.utils.vis_utils import plot_model

#進(jìn)行分字處理import osimport pandas as pdpath  = '/Users/wuzhengxiang/Documents/DataSets/TextCnn'os.chdir(path)text_all = pd.read_csv('text_all.csv')
data = text_alldata['text'] = text_all['text'].apply(lambda x: ' '.join(x))data.head()
x_train, x_test, y_train, y_test=\     train_test_split(data['text'],                       data['label'],                       test_size=0.2,                       random_state=42                      )print(x_train.shape, x_test.shape, y_train.shape, y_test.shape)

x_train[45]'\ufeff + C 名 看 P ， 是 威 哦 ?   ?'

MAX_NB_WORDS = 80000tokenizer=Tokenizer(num_words=MAX_NB_WORDS)tokenizer.fit_on_texts(x_train)

tokenizer.texts_to_sequences([x_train[38]])

train_sequences = tokenizer.texts_to_sequences(x_train)test_sequences  = tokenizer.texts_to_sequences(x_test)

MAX_LENGTH = 35padded_train_sequences = pad_sequences(train_sequences, maxlen=MAX_LENGTH)padded_test_sequences  = pad_sequences(test_sequences,  maxlen=MAX_LENGTH)
padded_train_sequencesarray([[  0,   0,   0, ...,   2,   8,   2],       [  0,   0,   0, ..., 387, 443,  16],       [  0,   0,   0, ...,  64,  59,  11],       ...,       [  0,   0,   0, ...,  27,  27,  27],       [  0,   0,   0, ...,   3,   9,  71],       [  0,   0,   0, ...,   1,  16,  16]], dtype=int32)padded_train_sequences.shape(15736, 35)
import numpy as np
def get_simple_rnn_model():    embedding_dim = 300    embedding_matrix = np.random.random((MAX_NB_WORDS, embedding_dim))    inp = Input(shape=(MAX_LENGTH, ))    x = Embedding(input_dim = MAX_NB_WORDS,                   output_dim = embedding_dim,                   input_length = MAX_LENGTH,                   weights=[embedding_matrix], trainable=True)(inp)    x = SpatialDropout1D(0.3)(x)    x = Bidirectional(GRU(100, return_sequences=True))(x)    avg_pool = GlobalAveragePooling1D()(x)    max_pool = GlobalMaxPooling1D()(x)    conc = concatenate([avg_pool, max_pool])    outp = Dense(1, activation="sigmoid")(conc)
    model = Model(inputs=inp, outputs=outp)    model.compile(loss='binary_crossentropy',                  optimizer='adam',                  metrics=['accuracy'])    return modelrnn_simple_model = get_simple_rnn_model()
plot_model(rnn_simple_model,            #to_file='C:\\Users\\伍正祥\\Desktop\\NLP任務(wù)實(shí)驗(yàn)\\rnn_simple_model.png',            show_shapes=True,            show_layer_names=True          )

filepath="weights-improvement-{epoch:02d}-{val_acc:.4f}.hdf5"checkpoint = ModelCheckpoint(filepath,                              monitor='val_acc',                              verbose=1,                              save_best_only=True,                             mode='max'                            )
batch_size = 256epochs = 2history = rnn_simple_model.fit(                   x = padded_train_sequences,                    y = y_train,                    validation_data = (padded_test_sequences, y_test),                    batch_size = 256,                    #callbacks=[checkpoint],                    epochs = 5,                    verbose=1           )
#best_rnn_simple_model = load_model('weights-improvement-01-0.8262.hdf5')
y_pred_rnn_simple = rnn_simple_model.predict(padded_test_sequences,                                                   verbose=1, batch_size=2048)
y_pred_rnn_simple = pd.DataFrame(y_pred_rnn_simple, columns=['prediction'])y_pred_rnn_simple['prediction'] = y_pred_rnn_simple['prediction'].map(lambda p: 1 if p >= 0.5 else 0)
#y_pred_rnn_simple.to_csv('./predictions/y_pred_rnn_simple.csv', index=False)#y_pred_rnn_simple = pd.read_csv('./predictions/y_pred_rnn_simple.csv')
print(accuracy_score(y_test, y_pred_rnn_simple))0.9669547534316217
df = pd.DataFrame({'text':x_test,'label':y_test,'pred':y_pred_rnn_simple})df[df['label']!=df['pred']]