Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 57717 (09)
- University: Sharif University of Technology
- Department: Civil Engineering
- Advisor(s): Nassiri, Habibollah
- Abstract:
- Pedestrian crashes pose a significant threat to public safety, accounting for a consid- erable proportion of road traffic fatalities and injuries worldwide. This study focuses on analyzing the severity of pedestrian crashes to provide insights that can help in developing effective countermeasures to enhance pedestrian safety. The research objectives include identifying key factors that impact the severity of pedestrian accidents or the frequency of severe accidents, comparing the performance of different statistical models in predicting accident severity and frequency, and rec- ommending the most effective model based on accuracy and other performance cri- teria. The models evaluated in this study include Multinomial Logistic and Poisson Regression, Bayesian updating methods, and Structural Equation Modeling (SEM). The methodology involves the collection and preparation of pedestrian crash data from urban areas in Pennsylvania, from 2014 to 2023. The data encompasses various attributes related to the crash, including environmental conditions, road characteris- tics, vehicle types, and pedestrian demographics. Each model is applied to the data to estimate the influence of these factors on crash severity. The results indicate that, overall, factors with p-values less than 0.05, such as ve- hicle speed, lighting conditions, and the use of drugs and alcohol have a significant impact on crash severity. Driver behavior factors have had the greatest influence, while road conditions and infrastructure have had the least impact on crash severity. Among environmental conditions, factors such as lighting and weather conditions and the time of the crash were identified as influential factors. Among pedestrian- related factors, the gender of the pedestrian was found to be insignificant. Finally, among vehicle-related factors, the number of vehicles and left-turn maneuvers had the greatest impact on crash severity. Of the evaluated models, SEM and Poisson Regression demonstrated superior perfor- mance in terms of prediction accuracy and managing complex variable interactions. SEM, in particular, provided valuable insights into the direct and indirect effects of various factors on pedestrian accident severity, making it a powerful tool for policy analysis. Conversely, the Bayesian updating method showed the lowest accuracy and the highest error, identifying it as the least effective model
- Keywords:
- Crash Severity ; Pedestrian Crashes ; Logistic Regression (LR)Analysis ; Poisson Count Regression ; Bayesian Updating ; Structural Equations Modeling ; Vulnerable Road Users
Digital Object List
-
محتواي کتاب
- view
Bookmark