基于貝葉斯網(wǎng)絡(luò)的高速公路交通事故研究
[Abstract]:Expressway has the characteristics of fast, closed, all-interchange traffic control mode, which provides a good condition for vehicle driving. The rapid development of freeway has greatly improved the traffic travel situation in our country. At the same time for our country's economic and social construction to provide favorable support. However, with the rapid increase of highway mileage, traffic accidents have also increased significantly, although the situation of expressway traffic accidents in China has been greatly improved in recent years. But the highway traffic accident still caused the serious harm to the people's life and property safety. Therefore, it is of great significance to analyze the expressway traffic accidents in order to improve the safety of expressway and to prevent and reduce the occurrence of traffic accidents. First of all, based on the research method of expressway accident distribution law, according to the existing expressway accident data and foreign highway traffic accident distribution law are compared and analyzed, the statistical analysis method is applied. The characteristics of expressway traffic accidents in China are obtained by analyzing the time distribution law, climate distribution law, accident form distribution law and vehicle type distribution law of expressway traffic accidents. Then, this paper analyzes the relationship between human, vehicle, road and environment in expressway traffic system operation and traffic accidents, and aims at the characteristics of high dimensional and nonlinear accident information. Support vector machine (SVM) algorithm is used to analyze the severity of accidents. The classification model is constructed by using support vector machine (SVM) algorithm, and the traffic accident data collected from expressway are studied, and the nonlinear SVM model is established according to the two classification method. Finally, the traffic accidents collected in this paper are classified according to the types of vehicles, which are divided into bus accidents, bus-truck accidents and truck accidents, and in the process of classification and modeling, The corresponding Bayesian network is established for the whole accident database and the three categories in order to explore the potential traffic accident law in different vehicle categories. Five indexes are introduced: correct rate, sensitivity, specificity, sensitivity and specificity harmonic mean (HMSS) index. The classification effect of the established Bayesian network model is evaluated by using ROC curve. Then, the results of Bayesian network structure learning show the different dependencies of each accident impact variable in different classification.
【學(xué)位授予單位】:北京交通大學(xué)
【學(xué)位級別】:碩士
【學(xué)位授予年份】:2017
【分類號】:U491.3
【參考文獻】
相關(guān)期刊論文 前9條
1 韋凌翔;陳紅;王龍飛;趙丹婷;邵海鵬;;誘發(fā)道路交通事故的關(guān)鍵因子分析方法研究[J];交通信息與安全;2015年01期
2 詹偉;呂慶;尚岳全;;高速公路隧道群交通事故灰色馬爾可夫預(yù)測[J];吉林大學(xué)學(xué)報(工學(xué)版);2014年01期
3 馬壯林;邵春福;李霞;;基于Logistic模型的公路隧道交通事故嚴重程度的影響因素[J];吉林大學(xué)學(xué)報(工學(xué)版);2010年02期
4 馬壯林;邵春福;李霞;;高速公路隧道交通事故嚴重程度的影響因素分析[J];北京交通大學(xué)學(xué)報;2009年06期
5 董國君;哈力木拉提·買買提;余輝;;基于RBF核的SVM核參數(shù)優(yōu)化算法[J];新疆大學(xué)學(xué)報(自然科學(xué)版);2009年03期
6 李世民;孫明玲;關(guān)宏志;;基于累積Logistic模型的交通事故嚴重程度預(yù)測模型[J];交通標準化;2009年Z1期
7 王興玲,李占斌;基于網(wǎng)格搜索的支持向量機核函數(shù)參數(shù)的確定[J];中國海洋大學(xué)學(xué)報(自然科學(xué)版);2005年05期
8 李文權(quán),王煒;交通事故的時間分布規(guī)律[J];中國安全科學(xué)學(xué)報;2005年04期
9 張孟奇,郭青;網(wǎng)格搜索法在送電線路鐵塔基礎(chǔ)計算機優(yōu)化設(shè)計中的應(yīng)用[J];電力建設(shè);1998年07期
相關(guān)博士學(xué)位論文 前3條
1 孫軼軒;基于數(shù)據(jù)挖掘的道路交通事故分析研究[D];北京交通大學(xué);2014年
2 張慧永;基于貝葉斯網(wǎng)絡(luò)的交通事故態(tài)勢研究[D];吉林大學(xué);2013年
3 李海生;支持向量機回歸算法與應(yīng)用研究[D];華南理工大學(xué);2005年
相關(guān)碩士學(xué)位論文 前5條
1 梁素芳;彎坡路段交通事故分析與預(yù)測[D];吉林大學(xué);2016年
2 程雨;基于貝葉斯網(wǎng)絡(luò)的列控系統(tǒng)故障診斷研究[D];北京交通大學(xué);2014年
3 徐晶;高速公路交通安全微觀評價方法及應(yīng)用研究[D];北京交通大學(xué);2011年
4 趙楊東;高速公路追尾事故成因分析及預(yù)防對策研究[D];哈爾濱工業(yè)大學(xué);2007年
5 沙愛敏;高速公路交通事故分析及預(yù)防對策研究[D];東南大學(xué);2006年
,本文編號:2235285
本文鏈接:http://sikaile.net/kejilunwen/daoluqiaoliang/2235285.html