网络服务器业务流量时间序列数据的片段式批量预测模型

打开文本图片集

中图分类号：TP391.4 文献标志码：A 文章编号： 1000-5013（2026）02-0235-12

Abstract：A time-frequency colaborative modeling traffic prediction method based on bidirectional gated recurrent unit （Bi-GRU） and filter neural network（FilterNet）is proposed，constructing a convolutional gated recurrent network-filter neural network （CG-FN） model. A fragmented batch prediction strategy is adopted， which employs a multi-input multi-output architecture to output complete future sequence fragments at once， fundamentally solving the error accumulation problem of traditional recursive prediction. By integrating convolutional neural network，Bi-GRU，and FilterNet，local， global，and time-frequency features are extrated，significantly improving the prediction accuracy of long-sequence traffic. Results show that the proposed model outperforms existing methods in prediction performance and generalization capability across multiple time scales and datasets，providing reliable support for server resource scheduling and network management.

Keywords：business trafic； long sequence prediction;time-frequency collaborative modeling；deep learning

近年来，云计算技术已演变为数字经济时代的关键基础架构域。（剩余18900字）