基于BP和SOM神经网络相结合的地震预测研究

Research on earthquake prediction based on BP and SOM neural network

摘要:

地震预测由于其产生原因的复杂性，一直是世界公认的难题。本文提出一种将多层前馈神经网络（BP网络）和自组织特征映射神经网络（SOM网络）相结合的方法并应用到地震震级的预测中，首先利用自组织特征映射神经网络对地震的原始数据进行聚类预处理，使具有内在规律的样本点集中在一起，之后利用BP神经网络对样本数据进行学习和预测，结果表明，相比线性回归预测模型和BP神经网络预测结果，增加SOM聚类处理过程能有效的减小预测误差。说明此方法可以有效的汇总出与地震关系密切的因素，也表明SOM对相关震级参数分类的有效性，对利用模糊预测方法来实现震级的预测是一种有效的辅助手段。

Abstract:

Because of the complexity of the causes of earthquake prediction，it has been recognized an aporia by all over the world. In this paper, a new method based on Back-Propagation neural network (BP) and Self-Organizing Feature Map neural network (SOM) is proposed，and applied to the prediction of earthquake magnitude. Firstly, Clustering of the original seismic data by using Self-Organizing Feature Map neural network,, which has the inherent law of the samples together, after using BP neural network to the sample data for learning and prediction, the experimental results show that compared with linear regression prediction model and BP neural network prediction results, the increase of SOM clustering process can effectively reduce the prediction error. It shows that this method can effectively summarize the factors which are closely related to earthquakes and SOM is effective for the classification of the relevant magnitude parameters, and it can be as an effective assistant method to predict the magnitude by using the fuzzy prediction method.