摘要
针对传统文本分类方法对配电网络缺陷文本分类性能不佳的问题,结合配电网缺陷文本数据特点,构建了基于改进卷积神经网络(CNN)的缺陷文本分类模型。首先,利用长短期记忆网络(LSTM)和门控循环单元(GRU),优化了配电网缺陷文本数据的词向量表示方式。然后,在CNN中引入注意力机制,提高了网络对重要文本的关注度,最后连接Softmax实现缺陷文本分类。基于实际算例对改进CNN模型与多种文本分类方法进行性能比较。结果表明,改进CNN模型对配电网缺陷文本分类性能优于其它文本分类方法。
The problem of poor text classification performance of traditional text classification methods for distribution network defects.Combining the characteristics of the defect text data of the distribution network,a defect text classification model based on an improved convolutional neural network(CNN)is constructed in this paper.Firstly,the word vector representation of the text data of the distribution network defects is optimized by using the long short-term memory(LSTM)and the gated recurrent unit(GRU).Then,an attention mechanism is introduced in CNN,which increased the network's attention to important texts,and finally connected Softmax to achieve the classification of defective texts.The performance comparison between the improved CNN model and various text classification methods is performed based on actual examples.The results show that the improved CNN model is better than other text classification methods in text classification of distribution network defects.
作者
党卫军
韩捷
薛艺为
DANG Wei-jun;HAN Jie;XUE Yi-wei(Guangzhou Power Supply Bureau Co.,Ltd.,Guangzhou 510000,China)
出处
《信息技术》
2020年第6期84-88,共5页
Information Technology
关键词
配电网缺陷文本
文本分类
卷积神经网络
注意力机制
distribution network defect text
text classification
convolutional neural network
attention mechanism
