CAJ | 학술논문

以发生交通事故的人、机、环境等多方面采集的调查数据为基础,解析突发性交通状态局部演变过程,并总结事发点通行能力的7类主要影响因素和3类评价指标.采用SPSS主成分分析与BP神经网络相结合的方法,对调查数据进行降维,并在保证数据丢失最小原则下,将七类影响因素提取为五类,简化神经网络拓扑结构,提高建模质量,通过BP神经网络模型调整离散输入量,以预测的评估值与实际评估值的误差最小为学习训练目标,求解最佳的连接强度权值与偏置值,得出事故因素与通行能力的定量关系.
이발생교통사고적인、궤、배경등다방면채집적조사수거위기출,해석돌발성교통상태국부연변과정,병총결사발점통행능력적7류주요영향인소화3류평개지표.채용SPSS주성분분석여BP신경망락상결합적방법,대조사수거진행강유,병재보증수거주실최소원칙하,장칠류영향인소제취위오류,간화신경망락탁복결구,제고건모질량,통과BP신경망락모형조정리산수입량,이예측적평고치여실제평고치적오차최소위학습훈련목표,구해최가적련접강도권치여편치치,득출사고인소여통행능력적정량관계.
Based on survey data of man, vehicles and environment involved in accidents, this study analyzes local evolution of traffic breakdown and summarizes influential factors of accident point capacity which include seven types of influential factors and three kinds of evaluating indicators. Dimensionality reduction of survey data is ful-filled by SPSS principal component analysis in combination with BP neural network. According to the minimum da-ta loss rule, seven types of influential factors are transformed to five principal components and neural network to-pology is simplified, thus improving the model quality. Adjustment of discrete inputs which use BP neural network model learning and training generates the optimal connection weights-W and bias values-B when the error be-tween predicted value and actual value reaches the minimum. The quantitative relationship between the factors of the accident and the capacity is thereby worked out.