Files in this item

FilesDescriptionFormat

application/pdf

application/pdfLI-THESIS-2020.pdf (2MB)
(no description provided)PDF

Description

Title:A machine learning model for vehicle crash type prediction
Author(s):Li, Xiyue
Advisor(s):Meidani, Hadi
Department / Program:Civil & Environmental Eng
Discipline:Civil Engineering
Degree Granting Institution:University of Illinois at Urbana-Champaign
Degree:M.S.
Genre:Thesis
Subject(s):Crash type, Crash prediction, Machine learning
Abstract:Travel safety research works include the studies of risk factor investigation, crash detection, and crash frequency prediction. However, the existing studies are focused on the macro level, paying little attention to the specific crash type. In this study, eXtreme Gradient Boosting (XGBoost) method is applied to predict the occurrence of different types of crashes. A two-layer model is proposed. The first layer is used to distinguish potential crashes from crash-free observations and the second layer is used for crash type recognition. The results show that the proposed model can detect the potential accident and identify the crash type successfully, with accuracy levels of over 99% and 62%, respectively. Besides the crash type prediction model, this study provides a detailed analysis of the impacts of different risk factors on different types of crashes. From the traffic management perspective, the results of this study can prepare traffic managers for the potential threatens well in advance. From travelers’ perspective, the results of this study can be used to warn travelers of potential dangers before the trip so that a better trip planning can be made as well as alert them of the potential dangers during the trip. All of these actions are important for the travel safety management and can help protect people’s life and property.
Issue Date:2020-05-15
Type:Thesis
URI:http://hdl.handle.net/2142/108066
Rights Information:Copyright 2020 Xiyue Li
Date Available in IDEALS:2020-08-26
Date Deposited:2020-05


This item appears in the following Collection(s)

Item Statistics