Please use this identifier to cite or link to this item:
Title: A Comprehensive Analysis of GCC Traffic Accidents with Machine Learning Techniques
Authors: Ahmed Maher Aiash 
Supervisor: Dr. Sharaf AlKheder
Degree Awarded: Master of Science Degree in: Civil Engineering
Keywords: Accidents, Severity, Risk Factors, Analysis, GCC
Issue Date: 2019
Publisher:  Kuwait university - college of graduate studies
Abstract: Traffic accidents are costing the nation more than a million lives a year alongside the monetary losses. In this thesis, the utilized data consisted of four different datasets. The first data set consisted of 5740 traffic accidents police reports that occurred in the UAE. A multinomial logit regression model was applied to determine the significant potential risk factors. The results showed that pedestrian, the unutilized seatbelt, roads that had four and more lanes, male casualty, 100 km/h speed limit or higher, and the casualty older than 60 years were found to be the factors that can increase the probability of being involved in a fatal traffic accident. In contrast, rear-end collisions and intersections had a lower likelihood of causing fatal injury. Then, the significant predictors were included in different mining techniques, including a neural network; CHAID tree; LSVM; and Bayesian network, to compare the performance of the applied methods and identify the normalized importance values for the significant independent variables. The second data set consisted of 9736 traffic accidents, that occurred from 2009 to 2010 in the UAE. These data were analyzed to provide an overview of the weather effects including fog, rain, dust, and fine weather conditions on traffic accidents types including hitting a pedestrian, accidents due to vehicle defects, and other types of crashes. Then, a multinomial logit regression model was applied to identify the correlations alongside a C5.0 tree model. Those four weather conditions were found to be significant in affecting traffic accidents types occurrence. Fine weather condition showed higher probabilities of having pedestrian traffic accidents. The third set of data consisted of 287,983 traffic accidents occurred in four-governorates. The types of traffic accidents were crashes, run-over, and rollover accidents. The fourth set of data comprised of 887,128 traffic violations were utilized to investigate the impact of gender on committing violations including passing red light, carelessness and risky behavior, and speeding violation. Then, a multinomial logit regression model was applied for both sets to identify the significant predictors and determine the correlations alongside utilizing a C5.0 tree. The results showed that both location and time were significant variables that influenced the occurring of certain types of accidents. Moreover, the results showed that male drivers had higher odds of having speeding violations and risky behavior compared to female drivers.
Appears in Programs:0620 Civil Engineering

Files in This Item:
File Description SizeFormat 
Thesis.pdf1,85 MBAdobe PDFView/Open    Request a copy
Show full item record

Page view(s)

Last Week
Last month
checked on Sep 29, 2020


checked on Sep 29, 2020

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.