地震震级预测的相关向量机模型

Relevance vector machine model for earthquake magnitude prediction

为了借助容易获取的地震相关因素间接预测地震震级,提出基于相关向量机(Relevance Vector Machine,RVM)方法的地震震级预测模型。通过样本学习建立地震震级与地震累积频度、累积释放能量、平均震级、b值、η值和相关区震级等6个主要影响因素之间的非线性映射关系,利用已知影响因素预测地震震级。结果表明:RVM模型预测结果均优于BP神经网络及SOM-BP神经网络预测结果;通过敏感因子分析比较各因素的敏感程度,b值和η值最为突出,在震级研究中应重点分析。综合分析,RVM模型具有精度高和离散性小等优点,对地震震级预测有较好的推广价值。

In order to predict earthquake magnitude using easily acquired earthquake correlation factors,an earthquake magnitude prediction model based on relevance vector machine(RVM)method is proposed.Through sample learning,the nonlinear mapping relationship between the earthquake magnitude and the cumulative frequency of earthquakes,the cumulative release energy,the average magnitude,the b value,theηvalue,and the correlation magnitude are established.Then,the known influencing factors can be used to predict the earthquake magnitude.The results show that the prediction results of RVM model are better than the prediction results of BP neural network and SOM-BP neural network;The sensitivity of each factor is compared by sensitive factor analysis,indicating that b value andηvalue are the most prominent factors and should be analyzed in the magnitude prediction.In general,the RVM model has the advantages of high precision and small dispersion,hence it has a good potential in the prediction of earthquake magnitude.