工业技术经济2025，Vol.44Issue(7)：111-123,13.DOI:10.3969/j.issn.1004-910X.2025.07.011
基于混频时序深度学习模型的汽车产业风险预测研究
Research on Risk Prediction of Automobile Industry Based on Mixed Frequency Time Series Deep Learning Model
摘要
Abstract
As a core link in modern industrial economic management,industrial risk prediction plays an irreplaceable role in ensuring coordinated economic development,optimizing industrial structure and scientifically formulating industrial development poli-cies.This paper proposes an innovative solution-Mixed-Frequency Temporal Fusion Dual Attention Network(MF-TF-DAN).What is particularly critical is that the MF-TF-DAN model innovatively introduces a dual attention mechanism,which starts from the two dimensions of time and features to deeply mine and evaluate the importance of the information processed by GRU and CNN.This paper conducted comprehensive and in-depth experimental verification on the industrial risk data set,including model compari-son experiments and ablation experiments with different prediction step sizes.Experimental results show that the MF-TF-DAN model performs significantly better than other comparison models in the mixing data prediction task.This result not only proves the scientific and effectiveness of the model design,but also brings new breakthroughs in the field of industrial risk prediction.This model provides industry managers with unprecedented accurate risk warning capabilities,allowing managers to have a deeper insight into market changes,identify and evaluate potential risks in advance,and thereby formulate more scientific and reasonable corpo-rate strategies and market response strategies.关键词
产业风险/汽车产业/深度学习/时间序列/混频数据/滑动窗口方法/电气能源/双重注意力机制Key words
industrial risk/automobile industry/deep learning/time series/mixed frequency data/sliding windows/e-lectrical energy/dual attention mechanism分类
经济学引用本文复制引用
刘洋,王广渠,韩立宁..基于混频时序深度学习模型的汽车产业风险预测研究[J].工业技术经济,2025,44(7):111-123,13.基金项目
教育部人文社会科学研究青年项目"基于社交机器人的突发公共卫生事件公众心理应激干预研究"(项目编号:22YJCZH114) （项目编号:22YJCZH114）
教育部人文社会科学研究青年项目"行业关联与金融风险:基于网络经济学方法的研究"(项目编号:19YJC790033) （项目编号:19YJC790033）
国家语委科研项目"领域数字化语言服务资源建设与关键技术研究"(项目编号:YB145-74). （项目编号:YB145-74）
