基于GCN-LSTM的频谱预测算法

Spectrum Prediction Algorithm Based on GCN-LSTM

无线频谱是一项重要的、难以再生的自然资源。在频谱数据中随着信道的动态变化,各个信道不能建模成规则的结构。由于卷积神经网络提取的是规则数据结构的相关性,没有考虑信道动态变化以及各个信道节点之间的相关性影响,基于此研究了基于图卷积神经网络(Graph Convolutional Network, GCN)和长短期记忆(Long Short-Term Memory, LSTM)网络结合的GCN-LSTM频谱预测模型,并且引入了注意力机制,仿真得到了GCN-LSTM在正确数据集和有一定错误数据的数据集上的预测性能和算法运行时间。结果表明在引入注意力机制后,GCN-LSTM预测模型的准确性和实时性都得到了提高。

Wireless spectrum is an important and hard-to-regenerate natural resource.Since convolutional neural network extracts correlation of regular data structure,dynamic changes of channel and the correlation between each channel node are not considered.Therefore,this paper studies a GCN-LSTM spectrum prediction model based on the combination of graph convolution neural network GCN and LSTM network,and introduces an attention mechanism.Simulation results show that the prediction performance and algorithm running time of GCN-LSTM on the correct dataset and the dataset with certain error data.Results show that the accuracy and real-time performance of GCN-LSTM prediction model are improved after introducing the attention mechanism.