發(fā)布時間：2023-06-03 03:12

　　道路交通安全狀態(tài)是復雜多因素協(xié)同作用的結果,道路交通事故背后的致因可以指導采取不同的措施來降低其危害。該研究評估了道路事故嚴重程度與天氣相關因子如何相關聯(lián)。天氣對高速公路交通安全的影響已成為道路交通安全部門日益關注的問題,如眾多與天氣有關的碰撞事故發(fā)生在潮濕路面和降雨狀態(tài)。因此,通過統(tǒng)計學和機器學習技術方法探索天氣相關變量與道路交通事故嚴重程度之間的關系非常重要。盡管以前的文獻中存在天氣對道路交通事故的影響,但需要構筑更加準確的模型,詳細分析解釋每個天氣因素的變化對碰撞事故嚴重程度的影響。本研究旨在基于公路安全信息系統(tǒng)(來自HSIS)事故數(shù)據(jù)集來分析和量化天氣相關因素(來自國家氣候數(shù)據(jù)中心數(shù)據(jù))對道路交通事故嚴重程度的影響。為了找到更好的模型擬合相關變量,研究選擇四個模型:order logit(OL)模型,決策樹模型,隨機森林模型和神經(jīng)網(wǎng)絡模型,分別在Stata和python中進行建模分析。在模型構建過程中,考慮了與天氣條件有關的七個主要因素,分別是氣溫,平均風速,日降水量,月降水量,年降水量,陣風和相對濕度。本文對比了統(tǒng)計學模型和機器學習模型的建模結果,通過靈敏度分析從機器學習模型...

【文章頁數(shù)】：76 頁

【學位級別】：碩士

【文章目錄】：
摘要
Abstract
Chapter1 Introduction
    1.0 Background
    1.1 An Explanation of Contributing Factors to Accidents Injury Severity
    1.2 Problem Statement of the Thesis:
    1.3 Research Aims and Objectives
    1.4 Organization and Summary of Thesis
Chapter2 Literature Review
    2.1 Introduction
        2.1.1 Impact of Weather on Vehicle Condition
        2.1.2 Impact of Weather on Road Condition
        2.1.3 Impact of Weather on Driver Behaviour
        2.1.4 Impact of Weather on Traffic Flow
    2.2 Accident Severity Models
    2.3 Limitations in Literature Review
Chapter3 Methodology
    3.1 Ordered Logistic Regression Model
    3.2 Artificial Neural Network or MLP
        3.2.1 Introduction:
        3.2.2 How ANN Algorithm Works
        3.2.3 Learning parameters
    3.3 Decision Tree
        3.3.1 Introduction
        3.3.2 The Decision Tree Learning Algorithm
        3.3.3 Deciding the“Best Attribute”
            Entropy
            Information Gain
            Gini Index
            Advantages of Decision Tree
            Disadvantages of Decision Tree
    3.4 Random Forest
        3.4.1 Introduction
        3.4.2 Feature Importance
        3.4.3 How does The Random Forests Algorithm work?
            Advantages of Random Forest
            Disadvantages of Random Forest
        3.4.4 Difference between Random Forests and Decision Trees
Chapter4 Data Collection and Preprocessing
    4.1 Data Collection
    4.2 Crash Severity Categories
    4.3 Variables Considered in the Study
    4.4 Data Description
    4.5 Data Preparation and Preprocessing
    4.6 Correlation Analysis
        Metrics for Evaluating Classification Models
Chapter5 Models Results and Analysis
    5.1 Introduction
    5.2 Statistical Analysis
    5.3 Machine Learning Models Results
        5.3.1 MLP Detailed Training Results:
            5.3.1.1 Deep Neural Network Training Results for Occupancy Dataset
            5.3.1.2 Deep Neural Network Results of Training on Vehicle Dataset
            5.3.1.3 Deep Neural Network Results of Training on Accident Dataset
        5.3.2 Random Forest Classifier
            Random Forest Results Summary
        5.3.3 Decision Tree Results Summary
    5.4 Comparison of ML Models Results
Chapter6 Results Discussion
    6.1 Impact of Weather Related Factors on Accident Severity
        6.1.1 Air Temperature
        6.1.2 Average wind speed
        6.1.3 Rainfall
        6.1.4 Humidity
        6.1.5 Wind Gust
    6.2 Summarization
Chapter7 Conclusion and Recommendations
    7.1 Conclusion
    7.2 Recommendations
Acknowledgements
References
List of Figures
List of Tables
List of Synonyms



本文編號：3828549