现代电子技术2025，Vol.48Issue(23)：69-74,6.DOI:10.16652/j.issn.1004-373x.2025.23.010
基于时序数据动态分析的舆情感知与演进监测技术
Public opinion perception and evolution monitoring technology based on dynamic analysis of time series data
摘要
Abstract
In view of the shortcomings of the existing public opinion supervision methods in contextual semantic understanding,time series dynamic modeling,and early warning,this paper proposes a dynamic analysis model that integrates improved language recognition and time series inference.In terms of public opinion perception,a BERT-TextCNN hybrid sentiment recognition algorithm is proposed to deeply capture global contextual semantics by improving BERT,and extract local key sentiment features by combining with TextCNN,so as to lay foundation for accurate perception.For the analysis of time series,a long short-term memory(LSTM)network model with attention mechanism(Att-LSTM)is constructed to model sentiment time series.This mechanism can weight key historical information dynamically,so that the turning points and trends of public opinion evolution can be captured accurately.The experimental results show that the accuracy rate of sentiment classification of the proposed model can reach 94.9%,which is significantly higher than that of the traditional baseline model.In the task of time series prediction,its root mean square error(RMSE)of the next three step prediction is as low as 0.084,which is obviously improved in the performance than that of the ARIMA model.This fully verifies the superiority and practical value of the proposed technical solution in terms of public opinion sensitivity,accuracy of evolutionary trend prediction,and early warning of sensitive information.关键词
长短期记忆网络/舆情感知/时序数据分析/情感识别/注意力机制/演进预测Key words
LSTM/public opinion perception/time series data analysis/sentiment recognition/attention mechanism/evolutionary prediction分类
信息技术与安全科学引用本文复制引用
郑小丽..基于时序数据动态分析的舆情感知与演进监测技术[J].现代电子技术,2025,48(23):69-74,6.基金项目
江西省高校人文社会科学研究专项项目:红色文化育人研究(HSWH25020) （HSWH25020）
