卷积神经网络CharCNN实现中文情感分类任务

import numpy as np
from keras.layers import Activation, Conv1D, Dense, Dropout, Embedding, Flatten, GlobalMaxPooling1D, Input
from keras.models import Model
class CharCNN:
    def __init__(self, max_seq_length, num_classes, vocab_size, embedding_dim=128, filter_sizes=(1, 2, 3), num_filters=128,
                 dropout_prob=0.5):
        self.max_seq_length = max_seq_length
        self.num_classes = num_classes
        self.vocab_size = vocab_size
        self.embedding_dim = embedding_dim
        self.filter_sizes = filter_sizes
        self.num_filters = num_filters
        self.dropout_prob = dropout_prob
    def build_model(self):
        inputs = Input(shape=(self.max_seq_length,), dtype='int32')
        x = Embedding(input_dim=self.vocab_size, output_dim=self.embedding_dim, input_length=self.max_seq_length)(inputs)
        conv_blocks = []
        for sz in self.filter_sizes:
            conv = Conv1D(filters=self.num_filters, kernel_size=sz, padding="valid", activation="relu", strides=1)(x)
            conv = GlobalMaxPooling1D()(conv)
            conv_blocks.append(conv)
        x = Concatenate()(conv_blocks) if len(conv_blocks) > 1 else conv_blocks[0]
        x = Dropout(self.dropout_prob)(x)
        x = Dense(256, activation='relu')(x)
        x = Dense(self.num_classes)(x)
        outputs = Activation("softmax")(x)
        model = Model(inputs=inputs, outputs=outputs)
        model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
        return model
    def train(self, X_train, y_train, X_test, y_test, batch_size=32, epochs=10):
        model = self.build_model()
        model.fit(X_train, y_train, batch_size=batch_size, epochs=epochs, validation_data=(X_test, y_test))
        return model
# 数据预处理
max_len = 100 # 最大文本长度
vocab_size = 5000 # 词汇表大小
embedding_dim = 50 # 嵌入维度
num_filters = 128 # 滤波器数量
filter_sizes = [3, 4, 5] # 滤波器尺寸
dropout_prob = 0.5 # Dropout概率
# 加载数据并进行预处理
from data_loader import DataLoader
data_loader = DataLoader(max_len,max_len,training=False)
X_train, y_train, X_test, y_test, word_index = data_loader.load_data()
vocab_size=len(word_index)
# 进行模型训练
char_cnn = CharCNN(max_seq_length=max_len, num_classes=3, vocab_size=vocab_size, embedding_dim=embedding_dim,
                     filter_sizes=filter_sizes, num_filters=num_filters, dropout_prob=dropout_prob)
model = char_cnn.train(X_train, y_train, X_test, y_test, batch_size=32, epochs=10)
# 进行预测
y_pred = model.predict(X_test)
y_pred = np.argmax(y_pred, axis=1)