摘要
地震预测在地球科学中是一项极具挑战性的任务,但由于地震数据呈现出非线性及复杂的时空特征,传统的预测方法难以有效处理。提出了一种结合双向长短期记忆网络(BiLSTM)与卷积门控循环单元(ConvGRU)的方法,应用于加州中部和北部的地震数据分析。该方法通过捕捉数据中的时空相关性,提升了模型的建模能力。实验结果显示,BiLSTM-ConvGRU模型在MSE和PSNR等评价指标上均显示出显著的优势,具有广阔的应用前景。
Earthquake prediction is a highly challenging task in Earth science,but due to the nonlinear and complex spatio-temporal characteristics of earthquake data,traditional prediction methods are difficult to effectively handle.A method combining bidirectional long short-term memory network(BiLSTM)and convolutional gated recurrent unit(ConvGRU)has been proposed for seismic data analysis in central and northern California.This method enhances the modeling capability of the model by capturing spatiotemporal correlations in the data.The experimental results show that the BiLSTM ConvGRU model exhibits significant advan-tages in evaluation metrics such as MSE and PSNR,and has broad application prospects.
作者
王思远
陈雨
Wang Siyuan;Chen Yu(College of Electronics and Information Engineering,Sichuan University,Chengdu 610065,China)
出处
《现代计算机》
2024年第21期14-19,共6页
Modern Computer
作者简介
王思远(1998-),男,浙江湖州人,在读硕士研究生,研究方向为深度学习在地震预测上的应用;通信作者:陈雨(1976-),男,四川成都人,博士,副教授,硕导,研究方向为地理信息系统,E-mail:3856704466@qq.com。
