基于混沌映射与飞行策略的短文本分类算法  

Research on Short Text Classification Algorithm Based on Chaotic Map and Flight Strategy

作　　者：苑津莎[1] 张瑾 张卫华[1] 班双双 YUAN Jinsha;ZHANG Jin;ZHANG Weihua;BAN Shuangshuang(Department of Electronic and Communication Engineering,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学电子与通信工程系,河北保定071003

出　　处：《电力科学与工程》2022年第4期17-23,共7页Electric Power Science and Engineering

摘　　要：针对极限学习机对文本分类所存在分类精度低的问题,使用结合变压器的双向编码器(BERT)和改进极限学习机(ELM)的短文本分类算法,提出引入Lévy飞行策略的混沌优化麻雀搜索算法(Lévy-CSSA)对ELM的权重与偏置进行寻优。该算法采用混沌映射使初始个体尽可能分布均匀,以增加初始种群的多样性,利用Lévy飞行搜索策略提高全局搜索能力。以电力客服工单为对象验证,结果表明,使用该方法可以更好地表达电力客服工单语义信息,相比所列举的其他经典模型,F_(1)值有明显提升,从而验证了模型的有效性。Aiming at the problem of low classification accuracy of text classification by extreme learning machine,a short text classification algorithm combining bidirectional encoder representation of transformer(BERT)and improved extreme learning machine(ELM)model is used,and chaotic optimization sparrow search algorithm with the introduction of Lévy light(Lévy-CSSA)is proposed to optimize the weight and bias of ELM.The algorithm uses chaotic mapping to make the initial individuals as uniform as possible to increase the diversity of the initial population,and uses the Lévy flight search strategy to improve the global search ability.Taking the power customer service work order as the verification object,the results show that the method in this paper can better express the semantic information of the power customer service.Compared with other classical models listed,the F_(1) value is significantly improved,which verifies the validity of the model.

关 键 词：BERT 麻雀搜索算法 ELM 混沌映射 电力客服工单 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]