现代电子技术2025，Vol.48Issue(22)：145-152,8.DOI:10.16652/j.issn.1004-373x.2025.22.024
基于NeuralProphet组合模型的云计算资源负载预测
Cloud computing resource load prediction based on NeuralProphet combination model
摘要
Abstract
The rapid development of cloud computing has increasingly intensified the load pressure on servers.Accurate load resource prediction has become a critical issue for resource allocation and server security in cloud centers.Existing single models struggle to capture global characteristics,while combined models often lack stability and interpretability when handling time series data.Therefore,a NeuralProphet-based combination model integrating convolutional neural networks(CNN),long short-term memory(LSTM)networks,and an Attention mechanism is proposed.NeuralProphet can decompose load time series data into trend,seasonal,and autoregressive components to enhance data stability and interpretability,so as to make the model capture global features and long-term dependencies more effectively.The dynamic weight allocation is conducted by means of the attention mechanism,and the key features that impact prediction outcomes are focused,further improving future load prediction accuracy.The experimental results on the Alibaba Cluster Data V2018 dataset show that the proposed combination model can outperform other deep learning models in terms of prediction accuracy and performance.In comparison with the traditional CNN-LSTM model,this model can improve the R2 score by 17.9％,and reduce the root mean square error(RMSE)by 73.6％,the mean absolute error(MAE)by 69.7％,and the symmetric mean absolute percentage error(sMAPE)by 65.3％.It has higher prediction accuracy and robustness,which is useful to improve the efficiency of cloud resource utilization.关键词
云计算/资源负载预测/NeuralProphet模型/卷积神经网络/长短期记忆网络/注意力机制/组合模型Key words
cloud computing/resource load prediction/NeuralProphet model/convolutional neural network/long short-term memory network/attention mechanism/combination model分类
电子信息工程引用本文复制引用
李好,谢晓兰,郭强..基于NeuralProphet组合模型的云计算资源负载预测[J].现代电子技术,2025,48(22):145-152,8.基金项目
国家自然科学基金资助项目:容器云资源管理优化方法研究(62262011) （62262011）
广西科技重大专项(桂科AA23062035) （桂科AA23062035）
广西重点研发计划项目(桂科AB23049001) （桂科AB23049001）
