摘要
随着电网设备运维的智能化发展,文本作为复杂的非结构化数据如何被有效利用已成为重要的问题。首先利用新词发现和传统分词方法对文本预处理,然后将字词特征映射至多维向量空间,最后基于特征融合构建了注意力机制优化卷积神经网络缺陷文本分类模型。算例分析表明,所提方法比传统深度学习方法提高了分类准确率,有更好的语义学习能力。
With intelligent development of operation and maintenance of power grid equipment,how to make effective use of text as complex unstructured data has become a major issue.Firstly,the text was preprocessed through new word discovery and traditional word segmentation method.Then,word features were mapped to a multidimensional vector space.Finally,based on feature fusion,a defective text classification model was constructed for the convolutional neural network for optimization of attention mechanism.Example analysis showed that the proposed approach had higher classification accuracy rate and better semantic learning ability than traditional deep learning.
作者
陈超
吴迪
唐昕
冯斌
张又文
郭创新
Chen Chao;Wu Di;Tang Xin;Feng Bin;Zhang Youwen;Guo Chuangxin(State Grid Zhejiang Pinghu Power Supply Co., Ltd., Pinghu Zhejiang 314200, China;State Grid Jiaxing Power Supply Co., Jiaxing Zhejiang 314033, China;College of Electrical Engineering, Zhejiang University, Hangzhou Zhejiang 310027, China)
出处
《电气自动化》
2021年第2期1-3,共3页
Electrical Automation
基金
基于大数据分析的输变配电设备状态跟踪与智能运检策略研究与应用项目(2019-LHKJ-014)。
关键词
特征融合
卷积神经网络
电力设备
文本挖掘
feature fusion
convolutional neural network
power equipment
text mining
