基于图卷积网络的基站用户数量预测

Prediction of Base Station Users Number Based on Graph Convolutional Network

移动网络用户预测建模有助于发现城市人群时空分布特征和实行合理资源调度策略。随着移动网络的快速发展,移动通信基站产生了大量手机信令信息,为数据驱动建模提供了支撑。传统的时序预测只考虑时序特征而忽略了空间上的关联。针对这一问题,提出一种基于图卷积网络(Graph Convolutional Network,GCN)的基站用户数量预测模型。利用GCN获取基站之间的空间关联特征,通过长短时记忆(Long Short-Term Memory,LSTM)网络对时序特征进行建模。通过进行对比实验和消融实验,证明该模型能够有效提取基站用户数量的时空特征,相对传统方法具有更优的性能。

Mobile user predictive modeling helps to discover spatial and temporal distribution characteristics of urban populations and implement reasonable resource scheduling strategies.With the rapid development of mobile networks,mobile communication base stations have generated a large amount of mobile signaling information,providing support for data-driven modeling.Traditional temporal prediction only considers temporal features and ignores spatial correlation.A model for predicting the number of base station users based on Graph Convolutional Network(GCN) is proposed to address this issue.Utilizing graph convolutional networks to obtain spatial correlation features between base stations,and modeling temporal features through Long Short-Term Memory(LSTM) network.The results of comparative experiments and ablation experiments show that the model can effectively extract the spatio-temporal characteristics of the number of base station users and have better performance compared to traditional methods.