基于粗糙集的BP神经网络在震例中的应用研究

Application of BP Neural Network Based on Rough Set in the Earthquake Case

采用《中国震例》作为数据源,通过初步整理分析和预处理,构建了较完备的震例研究样本集。尝试将粗糙集与BP神经网络相结合的方法引入到震例研究中,用基于粗糙集的属性约简算法从众多复杂的地震异常指标中筛选出对最终分类起决定作用的核心异常作为输入,震级作为输出,构建了泛化能力强的BP神经网络模型来模拟异常与地震之间的不确定关系。仿真测试结果表明:地震震级预测精度误差基本控制在-0.5～0.5级之间。

Firstly, using ＂China earthquake case＂ as data source, we built a fairly complete sample set for earthquake case study through preliminary analysis and pretreatment and introduced the combination of Rough Set and BP Neural Network to the earthquake case. Secondly, we selected the core abnormalities which plays a decisive role in final classification from a number of complex seismic anomaly indicators as the input by use of attribute reduction algorithm based on Rough Set, and took the discrete magnitude as the output. Furthermore, we built a generalized BP Neural Network model to simulate the uncertain relationship between the seismic anomaly and the earthquake. Finally, the result of simulation tests showed that the precision errors of earthquake magnitude prediction is between -0. 5 and 0. 5.