東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Traffic Crash Prediction Utilizing G...

Elrayah, Yassir E.

FindBook

Google Book

Amazon

博客來

Traffic Crash Prediction Utilizing Geospatial Analysis, Machine Learning, and Connected Vehicles' Basic Safety Messages.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Traffic Crash Prediction Utilizing Geospatial Analysis, Machine Learning, and Connected Vehicles' Basic Safety Messages./
作者:	Elrayah, Yassir E.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2020,
面頁冊數:	182 p.
附註:	Source: Dissertations Abstracts International, Volume: 81-11, Section: A.
Contained By:	Dissertations Abstracts International81-11A.
標題:	Transportation. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=27741692
ISBN:	9798645424817

Traffic Crash Prediction Utilizing Geospatial Analysis, Machine Learning, and Connected Vehicles' Basic Safety Messages.
Elrayah, Yassir E.

Traffic Crash Prediction Utilizing Geospatial Analysis, Machine Learning, and Connected Vehicles' Basic Safety Messages. - Ann Arbor : ProQuest Dissertations & Theses, 2020 - 182 p.

Source: Dissertations Abstracts International, Volume: 81-11, Section: A.

Thesis (Ph.D.)--Eastern Michigan University, 2020.

This item must not be sold to any third party vendors.

Traffic crash prediction using different techniques and methodologies is essential for road safety. While many researchers have evaluated road safety using historical crash data, there are a few traffic crash prediction studies that predict the likelihood of crashes using emerging technology such as connected vehicles. This technology enables a connected vehicle to broadcast driving behaviors, in the form of basic safety messages, at a frequency of 10 Hz to a nearby connected vehicle. Additionally, different geospatial analysis techniques and prediction methodologies, such as machine learning, can support the analysis, visualization, and prediction of the traffic crash frequency. The primary objective of this study was to predict the likelihood of crash incidents using connected vehicles' real-time driving behavior rather than historical crash data by developing and comparing different statistical and supervised machine learning predictive models. In addition, this study explored spatial and spatiotemporal patterns of crash incidents using geospatial analysis and unsupervised machine learning methodology. Moreover, it explored micro-level factors, such as road characteristics, and macro-level factors, such as population or points of interest density, which may impact traffic crash frequency. This study was based on a 3,000 connected-vehicle dataset, which was collected from the Safety Pilot Model Deployment Program in Ann Arbor, Michigan. The spatial analysis results using kernel density estimation showed that the city center and major road intersections were the most statistically significant high-risk locations. Additionally, the spatiotemporal analysis using emerging hotspot analysis showed that most of the city center roads had persistent hotspots, with some intensifying hotspots near road intersections and highway interchanges. Further, parametric and non-parametric correlation analysis and hierarchical regression models showed that some micro-level driving behavior variables and macro-level variables such as population, in relation to crash count, were statistically significant. Moreover, the results showed that the multinomial generalized mixed model with 91% prediction accuracy performed better than other statistical models. Finally, classification prediction, using supervised machine learning techniques, showed that both classification and regression tree and support vector machine algorithms, with multiple crash outcomes, performed well, with a high (94%) classification accuracy.

ISBN: 9798645424817Subjects--Topical Terms:

555912
Transportation.
Subjects--Index Terms:

Connected vehicles

Traffic Crash Prediction Utilizing Geospatial Analysis, Machine Learning, and Connected Vehicles' Basic Safety Messages.
LDR:03741nmm a2200373 4500 001 2269911
005 20200921070359.5
008 220629s2020 ||||||||||||||||| ||eng d
020 $a 9798645424817
035 $a (MiAaPQ)AAI27741692
035 $a AAI27741692
040 $a MiAaPQ $c MiAaPQ
100 1 $a Elrayah, Yassir E. $3 3547271
245 1 0 $a Traffic Crash Prediction Utilizing Geospatial Analysis, Machine Learning, and Connected Vehicles' Basic Safety Messages.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2020
300 $a 182 p.
500 $a Source: Dissertations Abstracts International, Volume: 81-11, Section: A.
500 $a Advisor: Xie, Yichun.
502 $a Thesis (Ph.D.)--Eastern Michigan University, 2020.
506 $a This item must not be sold to any third party vendors.
520 $a Traffic crash prediction using different techniques and methodologies is essential for road safety. While many researchers have evaluated road safety using historical crash data, there are a few traffic crash prediction studies that predict the likelihood of crashes using emerging technology such as connected vehicles. This technology enables a connected vehicle to broadcast driving behaviors, in the form of basic safety messages, at a frequency of 10 Hz to a nearby connected vehicle. Additionally, different geospatial analysis techniques and prediction methodologies, such as machine learning, can support the analysis, visualization, and prediction of the traffic crash frequency. The primary objective of this study was to predict the likelihood of crash incidents using connected vehicles' real-time driving behavior rather than historical crash data by developing and comparing different statistical and supervised machine learning predictive models. In addition, this study explored spatial and spatiotemporal patterns of crash incidents using geospatial analysis and unsupervised machine learning methodology. Moreover, it explored micro-level factors, such as road characteristics, and macro-level factors, such as population or points of interest density, which may impact traffic crash frequency. This study was based on a 3,000 connected-vehicle dataset, which was collected from the Safety Pilot Model Deployment Program in Ann Arbor, Michigan. The spatial analysis results using kernel density estimation showed that the city center and major road intersections were the most statistically significant high-risk locations. Additionally, the spatiotemporal analysis using emerging hotspot analysis showed that most of the city center roads had persistent hotspots, with some intensifying hotspots near road intersections and highway interchanges. Further, parametric and non-parametric correlation analysis and hierarchical regression models showed that some micro-level driving behavior variables and macro-level variables such as population, in relation to crash count, were statistically significant. Moreover, the results showed that the multinomial generalized mixed model with 91% prediction accuracy performed better than other statistical models. Finally, classification prediction, using supervised machine learning techniques, showed that both classification and regression tree and support vector machine algorithms, with multiple crash outcomes, performed well, with a high (94%) classification accuracy.
590 $a School code: 6456.
650 4 $a Transportation. $3 555912
650 4 $a Geographic information science. $3 3432445
653 $a Connected vehicles
653 $a Crash analysis
653 $a Geospatial
653 $a GIS
653 $a Machine learning
653 $a Transportation
690 $a 0370
690 $a 0709
710 2 $a Eastern Michigan University. $b Technology Studies. $3 2095790
773 0 $t Dissertations Abstracts International $g 81-11A.
790 $a 6456
791 $a Ph.D.
792 $a 2020
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=27741692