基于BP神经网络的交通事故预测方法研究  

Study of Traffic Accident Prediction Method Based on BP Neural Network

作　　者：唐阳山[1] 葛丽娜[1] 黄子龙[1] 杨培菲[1] 

机构地区：[1]辽宁工业大学汽车与交通工程学院,辽宁锦州121001

出　　处：《辽宁工业大学学报（自然科学版）》2016年第1期27-30,共4页Journal of Liaoning University of Technology(Natural Science Edition)

摘　　要：在分析道路交通事故影响因素的基础上,运用主成分分析法确定交通事故的9个主要影响指标,并以这9个指标为输入,以事故数、死亡人数、受伤人数及经济损失作为输出,建立交通事故BP神经网络模型。以1991-2011年的数据作为训练样本,2012年的数据作为检验样本,选取不同的训练函数,用matlab对网络进行训练,并对预测结果进行比较分析,结果表明BP神经网络对交通事故预测精度较高,且训练函数对预测精度有较大影响。On the basis of analyzing the factors of road traffic accidents, the principal component analysis（PCA） to determine the nine main impact indicators of traffic accidents is used, and nine indicators taken as input, the number of injured, deaths, and economic loss in accidents as output, the BP neural network model is established. To take the data from 1991 to 2011 as the training sample, and the data of 2012 as the test sample, different training functions to train network are chosen through matlab and the predicted results are analyzed and compared, the results show that the BP neural network for traffic accident prediction accuracy is higher, and training function has a great influence on training function of prediction accuracy.

关 键 词：事故预测 神经网络 主成分分析 训练函数 

分 类 号：U491[交通运输工程—交通运输规划与管理]