摘要
为解决文本分类中分类精度低的问题,提出一种混合优化的双模深度学习文本分类方法.该方法设计了一种混合优化算法,对深度学习模型进行权值调优,得到相关度高的特征和高性能文本分类结果.首先对文档进行预处理得到特征集合,设计了基于乌鸦搜索算法(CSA)和蝗虫优化算法(GOA)的混合优化算法,并使用双向门控循环单元(GRU)进行特征选择,得到具有上下文语义信息且相关的特征.最后,将最优特征输入到混合优化的深度置信网络(DBN)中得到文本分类结果.
In order to solve the problem of low classification accuracy in text classification,a dual-mode deep learning text classification method based on hybrid optimization is proposed.This method designs a hybrid optimization algorithm,optimizes the weights of the deep learning model,and obtains optimal features and high-performance text classification results.First,the document is preprocessed to obtain the feature set,and a hybrid optimization algorithm based on the crow search algorithm and the grasshopper optimization algorithm is designed.The hybrid optimized Bi-GRU is used to select the optimal features,and the highly relevant features with context semantic information are obtained.Finally,the optimal features are input into the DBN with hybrid optimized weights to obtain the text classification results.
作者
吴绪玲
WU Xuling(School of Information Engineering,Hope College,Southwest Jiaotong University,Chengdu 610400,China)
出处
《西南大学学报(自然科学版)》
CAS
CSCD
北大核心
2022年第11期234-242,共9页
Journal of Southwest University(Natural Science Edition)
基金
四川省电子商务与现代物流研究中心项目(DSWL21-27)
成都市哲学社会科学重点研究基地课题(2019002).
关键词
文本分类
混合优化算法
深度学习
双向门控循环单元
深度置信网络
text classification
crow search algorithm
grasshopper optimization algorithm
bi-gated recurrent unit
deep confidence network