基于RIME-CNN-LSTM-Attention的频谱占用预测研究
首发时间:2024-06-27
摘要:频谱占用预测在动态频谱管理中具有重要意义。本文提出了一种基于霜冰优化算法(RIME)、卷积神经网络(CNN)、长短期记忆网络(LSTM)及注意力机制(Attention)的频谱占用预测模型(RIME-CNN-LSTM-Attention),并对其性能进行了系统评估。通过与LSTM-Attention、CNN-Attention、CNN-LSTM-Attention三种常用模型进行对比分析,本文采用了均方误差(MSE)、平均绝对百分比误差(MAPE)、解释方差(EV)及决定系数(R2)对各模型的预测性能进行评估。实验结果表明,RIME-CNN-LSTM-Attention模型在所有评估指标上均表现出更高的预测精度和稳定性,充分验证了其在频谱占用预测任务中的优越性。本文的研究为动态频谱管理提供了有效的技术支持,具有重要的理论价值和实际应用前景。
关键词: 无线通信 频谱占用预测 霜冰优化算法 卷积神经网络 长短期记忆网络 注意力机制
For information in English, please click here
Spectrum Occupancy Prediction Based on RIME-CNN-LSTM-Attention
Abstract:Spectrum occupancy prediction plays a crucial role in dynamic spectrum management. This paper proposes a spectrum occupancy prediction model based on the Rime Optimization Algorithm(RIME), Convolutional Neural Network (CNN), Long Short-Term Memory Network (LSTM), and Attention Mechanism, referred to as the RIME-CNN-LSTM-Attention model. The performance of this model is systematically evaluated and compared with three commonly used models: LSTM-Attention, CNN-Attention, and CNN-LSTM-Attention. The evaluation metrics used include Mean Squared Error (MSE), Mean Absolute Percentage Error (MAPE), Explained Variance (EV), and the Coefficient of Determination (R2). Experimental results indicate that the RIME-CNN-LSTM-Attention model demonstrates higher prediction accuracy and stability across all evaluation metrics, thereby validating its superiority in spectrum occupancy prediction tasks. This research provides effective technical support for dynamic spectrum management and holds significant theoretical value and practical application prospects.
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
基于RIME-CNN-LSTM-Attention的频谱占用预测研究
评论
全部评论