基于机器学习的强震动监测环境抗干扰方法对比研究  

RESEARCH ON ANTI-JAMMING TECHNOLOGY OF STRONG MOTION BASED ON MACHINE LEARNING

作　　者：庞聪[1,2] 江勇 廖成旺 吴涛[1,2] 丁炜 PANG Cong;JIANG Yong;LIAO Cheng-wang;WU Tao;DING Wei(Key Laboratory of Earthquake Geodesy,Institute of Seismology,CEA,Wuhan 430071,Hubei,China;Hubei Key Laboratory of Earthquake Early Warning,Hubei Earthquake Agency,Wuhan 430071,Hubei,China)

机构地区：[1]中国地震局地震研究所,中国地震局地震大地测量重点实验室,湖北武汉430071 [2]湖北省地震局,地震预警湖北省重点实验室,湖北武汉430071

出　　处：《内陆地震》2020年第2期119-124,共6页Inland Earthquake

基　　金：中国地震局地震研究所所长基金(IS201856290)。

摘　　要：引入机器学习中的决策树、随机森林、AdaBoost集成学习等方法,分别按照训练比例为10%~90%、变化率为10%的试验方式,从识别准确度、算法执行时间、异常数等角度出发,对比分析强震动数据抗干扰算法在不同样本训练量、不同验证数据量下的识别效果、执行效率及算法稳健性。实验结果表明,AdaBoost集成学习的识别效果与稳定性最好,但是算法效率较差,决策树的算法稳定性较差,但是效率较高。综合算法性能来看,随机森林的应用前景较大,具有一定实用价值。This paper introduces decision tree,random forest,AdaBoost integrated learning and other methods in machine learning.According to the test mode with training proportion of 10%~90%and change rate of 10%,and from the perspective of recognition accuracy,algorithm execution time,abnormal number,etc.,it compares and analyzes the recognition effect,execution efficiency and performance of strong motion data anti interference algorithm under different sample training amount and different verification data amount Algorithm robustness.The experimental results show that AdaBoost ensemble learning has the best recognition effect,but the algorithm efficiency is poor,the algorithm stability of decision tree is poor,while the efficiency is high.According to the performance of the algorithm,the application prospect of random forest is great,and it has a certain practical value.

关 键 词：强震动监测 抗干扰 ADABOOST 决策树 随机森林 

分 类 号：P315.7[天文地球—地震学]