基于灰色BP神经网络组合模型的水上交通事故数预测  

Integrated model for forecasting waterway traffic accidents based on the Gray-BP neural network

作　　者：范中洲[1] 赵羿 周宁 赵冲 张文烨 FAN Zhong-zhou;ZHAO Yi;ZHOU Ning;ZHAO Chong;ZHANG Wen-ye(Navigation College,Dalian Maritime University,Dalian 116026,Liaoning,China)

机构地区：[1]大连海事大学航海学院,辽宁大连116026

出　　处：《安全与环境学报》2020年第3期857-861,共5页Journal of Safety and Environment

基　　金：国家自然科学基金项目(51179019)。

摘　　要：为了提高水上交通事故数的预测精度,在灰色预测模型与BP神经网络模型相组合的基础上,建立了灰色BP神经网络组合预测模型。以全国水上交通事故数、机动船数、驳船数、水上运输就业人数、水路货运量和水路客运量的数据作为样本数据,进行试验计算,并将组合预测模型的预测结果分别与灰色预测模型和BP神经网络预测模型的预测结果进行了对比。试验结果表明,相较灰色预测模型和BP神经网络预测模型灰色BP神经网络组合预测模型的误差更小、预测精度更高,且具有良好的稳定性。The paper is aimed at establishing an integrated model for forecasting waterway transportation and traffic accidents based on the Gray-BP neural network model. As compared with the traditional gray prediction model and BP neural network prediction model,the said forecasting model takes the 3 major factors of waterway traffic and transportation accidents and features of the linear time series into account,with the 4 following steps included in it. To achieve the purpose,it is necessary,first of all,to establish a gray prediction model to predict the original time series of the waterway traffic accident number and get the prediction sequence. And,next,the gray prediction sequence has to be gained from the original time series of the waterway traffic accident number to gain the residual sequence. And,thirdly,the residual sequence has to be brought into the modified BP neural network prediction model to get the residual correction sequence. And,last of all,that is,the last step,the gray prediction model prediction sequence has to be aided to the modified residual sequence of the BP neural network so as to obtain the Gray-BP neural network prediction outcome. Thus,it can be seen that the paper has taken the historical data of China’s national number of the waterway traffic accidents as the reference sampling data from 2001 to 2014,including the vessel numbers,barge numbers,crew number sampling,the total cargo transportation and passenger number transportation sum-total as a reference basis. The results of our prediction have been built up on the basis of the integrated prediction model with the results of the gray model prediction and that of the BP neural network prediction model individually and separately. Meanwhile,the paper has also analyzed the error and stability of the gray prediction and the root mean square error (RMSE),the mean absolute error (MAE) by using the BP neural network prediction model and the gray BP integrated prediction model. Thus,the final results have shown that the Gray-BP prediction mode

关 键 词：安全工程 水上交通事故 事故预测 组合预测模型 

分 类 号：X951[环境科学与工程—安全科学]