首页|基于Transformer的时间序列预测研究

基于Transformer的时间序列预测研究

扫码查看
文章探究Transformer结构在时间序列预测领域的性能,并讨论目前的工作与方法,分析改进方向.通过改进Transformer结构应用于长时间序列预测,降低计算复杂度和时间成本,同时提高预测准度.分别进行单变量和多变量预测,并与不同的基准模型进行对比得出实验结果,发现基于Transformer结构改进的模型在时间序列预测方面效果较好.
Research on Time Series Prediction Based on transformer
The article explores the performance of transformer structure in the field of time series prediction,discusses the current work and methods,and analyzes the direction of improvement.The improved transformer structure for long time series forecasting reduces computational complexity and time cost,while improving prediction accuracy.Then,univariate and multivariate predictions are made respectively and compared with different benchmark models.The experimental results show that the improved model based on transformer structure can achieve better results in time series prediction.

transformertime seriesself-attention mechanismforecast

潘美财、吴楠

展开 >

南宁师范大学计算机与信息工程学院,广西南宁 530100

Transformer 时间序列 自注意力机制 预测

广西壮族自治区研究生教育创新计划

YCSW2023437

2024

信息与电脑
北京电子控股有限责任公司

信息与电脑

影响因子:1.143
ISSN:1003-9767
年,卷(期):2024.36(2)
  • 10