基于PCA-ELM的地震死亡人数评估

地震往往会在短时间内造成巨大的人员伤亡和财产损失,为了震后救援工作能够快速高效的展开,建立了主成分分析(Principal Components Analysis,PCA)和极限学习机(Extreme Learning Machine,ELM)相结合的地震死亡人数预测模型。综合考虑影响地震死亡人数的多种因素,选取震级、震源深度、震中烈度、抗震设防烈度、人口密度、发生时间和预报水平7项主要因素作为评价指标。首先,利用主成分分析对原始变量进行降维处理,并计算出主成分得分,作为ELM的输入;其次,对构建的ELM地震死亡人数预测模型进行训练;最后,对选取的32组地震样本进行算例仿真,并与未经PCA处理的ELM算法和BP神经网络算法进行对比。结果表明,基于主成分分析的极限学习机算法对地震死亡人数预测具有较高的预测精度,验证效果良好。

Evaluation of the earthquake death toll based on the PCA-ELM analysis

The present paper intends to establish a forecast or prediction model of earthquake death toll based on the principal components analysis(PCA)and the extreme learning machine(ELM)so as to carry out the rescue work promptly and efficiently after the earthquake takes place.As is known,earthquakes often bring about huge death casualty and property losses in a short period of time.Taking into account a variety of factors affecting the number of earthquake death toll,we have taken 7 major factors into account as the evaluation indicators,including the magnitude,the quake source depth,the epicenter intensity,the seismic fortification intensity,the population density,the occurrence time and the forecast accuracy level.Among the abovementioned influential factors,it is necessary first of all to analyze the principal components to reduce the dimensionality of the original variables to eliminate the interaction between the influential factors and reject data redundancy so as to obtain the characteristic features,the contribution rates of all the principal components,and clarify the cumulative contribution rates,and the principal component score as an input as an ELM model.And,then,it is also necessary to build up the ELM earthquake death population prediction model.Among the 32 seismic samples,we have chosen 27 samples randomly as the input data for training and the principal component score.And,so,the training results can be processed by the 10 logarithms,with the number of actual death toll compared with that of the predicted one,and the accuracy of the comparison measured and tested from the point of view of the macro perspective.And,thus,finally,we have chosen 5 groups of seismic samples to test the model to be trained.To further verify the validity of the PCA–ELM model in the prediction of the earthquake death toll,the algorithm has been used to compare with the ELM algorithm with no need for PCA processing and the BP neural network algorithm.And,thus,in contrast to the results gained,it would be possible to see and verify the prediction accuracy of the 3 models.The results of above all investigations and calculations demonstrate that the extreme learning machine algorithm based on the principal component analysis enjoys high prediction accuracy for the earthquake death toll prediction.It can help to predict the number of the earthquake death toll well enough so as to provide a more accurate prediction for the earthquake death toll at a highly precise level.