基于地震观测数据的机器学习地震预测研究综述

(1.中国地震局地震预测研究所,北京 100036; 2.中国地震台网中心,北京 100045)

地震预测; 机器学习; 特征提取; 模型评价

Research Progress in Field of Earthquake Prediction by Machine Learning Based on Seismic Data
WANG Jinhong1,JIANG Haikun2

(1.Institute of Earthquake Forecasting,China Earthquake Administration,Beijing 100036,China)(2.China Earthquake Networks Center,Beijing 100045,China)

earthquake prediction; machine learning; feature extraction; model evaluation

DOI: 10.20015/j.cnki.ISSN1000-0666.2023.0022

备注

机器学习突出的隐式特征提取和复杂任务处理能力正推动着地震预测科学的发展,为系统了解机器学习技术在地震预测领域的发展现状,从指定时空窗的地震震级预测、发震位置和发震时间估计三方面,综述了国内外机器学习在地震预测领域中的应用,其中在震级预测问题上AI应用最为广泛; 总结了机器学习地震预测的主要特征参数、模型和评价相关问题,从多种评价机制中探索地震活动性参数对地震预测结果的影响,并对地震预测领域存在的问题进行初步讨论和展望。在可预见的未来,AI技术的引入和应用领域的拓展,有可能引领地震预测领域的持续发展。 关键词:地震预测; 机器学习; 特征提取; 模型评价
Machine learning's prominent implicit feature extraction and complex task processing capabilities are driving the science of earthquake prediction.In order to systematically understand the development status of machine learning technology in the field of earthquake prediction,this paper focuses on the application of machine learning in the field of earthquake prediction at home and abroad in recent years from the three aspects of earthquake magnitude prediction,earthquake location and earthquake occurrence time estimation in a specified time-space window,among which AI is the most widely used in earthquake magnitude prediction.In addition,this paper summarizes the main characteristic parameters,models and evaluation related issues of machine learning earthquake prediction,and explores the influence of seismicity parameters on earthquake prediction results from various evaluation mechanisms.Finally,a preliminary discussion and outlook on the problems existing in the field of earthquake prediction will be conducted.In the foreseeable future,the introduction of AI technology and the expansion of application fields are likely to lead the continuous development of the field of earthquake prediction.