基于Transformer的网络流量预测研究
Research on network traffic prediction based on Transformer
田爱宝 1魏娇娇 2肖军弼2
作者信息
- 1. 中国石油大学(华东)信息化建设处,山东青岛 266580
- 2. 中国石油大学(华东)计算机科学与技术学院,山东青岛 266580
- 折叠
摘要
网络流量预测是网络流量分析领域中亟待解决的关键任务之一.现基于机器学习的预测方法大多忽略了流量的长相关性,并且处理大量数据时耗时长.针对以上问题,文中将Transformer用于网络流量预测,通过多头注意力机制捕获流量的远程序列关系,学习流量的全局依赖关系.实验结果表明,该方法可以提高预测精度,并能有效降低训练时间.
Abstract
Network traffic prediction is one of the key tasks to be solved urgently in the field of network traf-fic analysis.Most of the current prediction methods based on machine learning ignore the long correlation of traffic and take a long time to process large amounts of data.In response to the above problems,the study uses Transformer for network traffic prediction,captures the long-range sequence relationship of traffic through a multi-head attention mechanism,and learns the global dependency of traffic.The experiment re-sults show that this method can improve the prediction accuracy and effectively reduce the training time.
关键词
流量预测/Transformer/深度学习/注意力机制/特征提取Key words
traffic prediction/Transformer/Deep Learning/attention mechanism/feature extraction引用本文复制引用
出版年
2024
NETL
NSTL
万方数据