基于数据分类的高速铁路运营事故预测方法*

高宁波1胡启洲1▲张兵2郑丽媛1

(南京理工大学自动化学院南京 210094；华东交通大学土木建筑学院南昌 330013)

摘要高速铁路运营事故预测方法是度量铁路安全管理水平的重要指标。为提高高速铁路的安全运营水平，引入工业数据分类方法，分析反向传播(BP)神经网络和灰色模型在高速铁路安全运营事故预测过程中的适应性。首先，运用事故次数、事故联动系数、月均事故率3个参数对高速铁路安全运营水平进行度量；然后，根据工业数据分类方法判别高速铁路运营事故数据属于块状型，据此建立反向传播(BP)神经网络运营事故预测模型；针对运营事故数据具有波动大的特点，利用均值聚类方法建立K-GM(1,3)预测模型。以近年来高速铁路运营事故数据为样本对模型进行训练和分析，结果表明：BP神经网络、K-GM(1,3)、GM(1,3)预测模型的预测误差分别为8.92%,13.68%,345.25%，BP神经网络在高速铁路安全运营事故预测过程中的适应性要优于灰度模型。

关键词交通安全；事故预测；数据分类；高速铁路； 神经网络；灰色模型

中图分类号：X913.3文献标志码：A

收稿日期：2014-07-22修回日期：2014-09-27

基金项目*国家自然科学(批准号:51178157)、中央高校基本科研业务专项

作者简介：第一高宁波(1990-)，硕士研究生.研究方向：交通拥堵，交通安全.E-mail:916747600@qq.com

通讯作者：▲胡启洲(1975-)，博士，副教授.研究方向：交通管理与控制.E-mail:qizhouhu@163.com

An Accident Prediction Model of High-Speed Railway

Operation Based on Data Classification

GAO Ningbo1HU Qizhou1ZHANG Bing2ZHENG Liyuan1

(1.SchoolofAutomation,NanjingUniversityof

science&Technology,Nanjing2100984,China;

2.EastChinaJiaoTongUniversity，Nanchang330013，China)

Abstract:Accident prediction models for high-speed railway are important measurements for safety management. In order to improve the safety of high-speed railway, the capability of back propagation (BP) neural network and grey model in the process of estimating accidents of high-speed railway are analyzed. Firstly, the number, the coefficients of causal factors and average number of accidents per month are used to assess the safety management of high-speed railway in this paper. Then, the accident prediction models for high-speed railway is developed based on BP Neural Network and GM(1,3). Due to the fact that the accident data is quite unstable, combined with k-means clustering method, the K-GM(1,3) models are developed. The accident data of the high-speed railway collected in recent years are used to train and test the models developed. Study results show that the mean relative error of BP Neural Network, K-GM(1,3) and GM(1,3) models are 8.92%, 13.68% and 345.25% respectively. It is found that the BP neural network model perform much better than the grey model.

Key words:Transportation safety; Accident prediction; Data classification; High-speed railway; Neural Network; Grey Model

0引言

2013年国内高速铁路运营里程达到11 028 km，分别是德国和日本的11.6倍和4.6倍，跃居世界第1位。同时，随着运营速度的提升，高速铁路运营安全问题已成为社会各界关注的热点问题。……