基于機(jī)器學(xué)習(xí)與統(tǒng)計(jì)學(xué)模型的道路事故嚴(yán)重程度預(yù)測(cè)模型效果評(píng)價(jià)
發(fā)布時(shí)間：2024-05-16 21:56
道路交通事故數(shù)據(jù)分析對(duì)于交通安全有著重要意義。事故分析的重要性在于可以揭示導(dǎo)致事故的不同類型因素的影響。道路事故風(fēng)險(xiǎn)模型的預(yù)測(cè)準(zhǔn)確性需要不斷提高。數(shù)據(jù)挖掘方法可以用于道路交通事故數(shù)據(jù)分析。其中,統(tǒng)計(jì)學(xué)模型OP、MNL等以及機(jī)器學(xué)習(xí)模型CART,SVM,KNN,GNB和RF等均可用于道路交通事故的數(shù)據(jù)集分析。這給我們提供了去研究更加準(zhǔn)確模型的機(jī)會(huì)。本文對(duì)比了基于具有不同建模邏輯的各種機(jī)器學(xué)習(xí)和統(tǒng)計(jì)學(xué)模型在道路事故損害程度預(yù)測(cè)中的精確度�；谙愀鄄煌貐^(qū)委員會(huì)收集的道路事故數(shù)據(jù),將這些模型用于預(yù)測(cè)與各道路事故程度等級(jí)相對(duì)應(yīng)的損害嚴(yán)重程度。本文計(jì)算并比較每個(gè)模型在測(cè)試數(shù)據(jù)集上的預(yù)測(cè)準(zhǔn)確性,然后進(jìn)行靈敏度分析以推斷解釋變量對(duì)道路事故嚴(yán)重程度判斷的重要性。并對(duì)比了OP和MNL統(tǒng)計(jì)學(xué)模型對(duì)于變量影響的估計(jì)。從靈敏度分析中,我們可以獲得五個(gè)選定的機(jī)器學(xué)習(xí)模型對(duì)于碰撞事故嚴(yán)重程度的影響大小。結(jié)果表明,盡管機(jī)器學(xué)習(xí)模型的方法存在過(guò)度擬合的問(wèn)題,但其相比統(tǒng)計(jì)學(xué)模型的方法具有更高的預(yù)測(cè)準(zhǔn)確性。RF,GNB,KNN,SVM和CART的致命事故分類準(zhǔn)確率分別為82.77%,55.53%,82.82%,77.93...
【文章頁(yè)數(shù)】：85 頁(yè)
【學(xué)位級(jí)別】：碩士
【文章目錄】：
摘要
Abstract
Chapter 1 Introduction
1.1 Background
1.2 Statistics Summary of Crash Contributing Factors
1.3 Problem Statement and Intention of This Thesis
1.4 Research Aim and Objectives
1.5 Outline of the thesis
Chapter 2 Literature Survey
2.1 Introduction
2.2 Status of Road Safety
2.2.1 Status of Road Safety around the World
2.4 Literature Survey of Statistical Models for Crash Injury Severity
2.5 Literature Survey of Machine Learning Models for Crash Injury Severity
2.6 Summarization and Limitations
Chapter 3 Methodology for Crash Modeling
3.1 Introduction
3.2 The Design of Methodology
3.3 Statistical Models
3.3.1 Ordered Probit Regression Model
3.3.2 Checking for Multi-Collinearity
3.3.3 Multinomial Logistic Model(MNLM)Design
3.4 Machine Learning Models
3.4.1 Classification and Adaptive Regression Trees(CART)
3.4.1.1 Data Set Portioning
3.4.1.2 Choose Cost Function and Training Model
3.4.1.3 Decision Tree Algorithm Advantages and Disadvantages
3.4.2 Support Vector Machine
3.4.3 Naive Bayes Classifier
3.4.3.1 What is Bayes Theorem?
3.4.3.2 Types of Naive Bayes Algorithm
3.4.3.3 Representation Used By Naive Bayes Models
3.4.3.4 Make Predictions with a Naive Bayes Model
3.4.3.5 Na?ve Bayes Algorithm Advantages and Disadvantages
3.4.4 K-Nearest Neighbors– Classification
3.4.4.1 Algorithm
3.4.4.2 K-NN Algorithm Advantages and Disadvantages
3.4.5 Random Forest
3.4.5.1 How does The Random Forests Algorithm work?
3.4.5.2 Feature Importance
3.4.5.3 Random Forest Algorithm Advantages and Disadvantages
Chapter 4 Crash Data Collection and Data Description
4.1 Introduction
4.1.1 Hong Kong Transportation Department Accident Data
4.1.2 Variables Considered In the Study
4.2 Data Preparation
4.2.1 Based on Accident
4.2.2 Based on Vehicle
4.2.3 Based on casualty
4.3 Data Pre-Processing
4.3.1 Missing Data Treatment
4.3.2 Data Normalization
4.4 Estimation of Accuracy in Classification
4.5 Models Selection by Performance Evaluation
Chapter 5 Data Analysis and Modeling Results
5.1 Statistical Models Results
5.2 Machine Learning Models Analysis Results
5.3 Experiments and Results
5.3.1 CART Experimental Results
5.3.2 Support Vector Machine Results
5.3.3 K-Nearest Neighbor Results
5.3.4 Gaussian Na?ve Bayes Results
5.3.5 Random Forest Results
5.4 Results Comparison of Machine Learning Models
5.5 Summary
Chapter 6 Discussion of Findings
6.1 Sensitivity Analysis
6.2 Comparison of Variable Impact on Crash Severity from ML Models
6.3 Summary
Chapter 7 Conclusion and Recommendations
7.1 Conclusion
7.2 Recommendations
References
Acknowledgements
List of Figures
List of Tables
List of Acronyms
本文編號(hào)：3974965
【文章頁(yè)數(shù)】：85 頁(yè)
【學(xué)位級(jí)別】：碩士
【文章目錄】：
摘要
Abstract
Chapter 1 Introduction
1.1 Background
1.2 Statistics Summary of Crash Contributing Factors
1.3 Problem Statement and Intention of This Thesis
1.4 Research Aim and Objectives
1.5 Outline of the thesis
Chapter 2 Literature Survey
2.1 Introduction
2.2 Status of Road Safety
2.2.1 Status of Road Safety around the World
2.4 Literature Survey of Statistical Models for Crash Injury Severity
2.5 Literature Survey of Machine Learning Models for Crash Injury Severity
2.6 Summarization and Limitations
Chapter 3 Methodology for Crash Modeling
3.1 Introduction
3.2 The Design of Methodology
3.3 Statistical Models
3.3.1 Ordered Probit Regression Model
3.3.2 Checking for Multi-Collinearity
3.3.3 Multinomial Logistic Model(MNLM)Design
3.4 Machine Learning Models
3.4.1 Classification and Adaptive Regression Trees(CART)
3.4.1.1 Data Set Portioning
3.4.1.2 Choose Cost Function and Training Model
3.4.1.3 Decision Tree Algorithm Advantages and Disadvantages
3.4.2 Support Vector Machine
3.4.3 Naive Bayes Classifier
3.4.3.1 What is Bayes Theorem?
3.4.3.2 Types of Naive Bayes Algorithm
3.4.3.3 Representation Used By Naive Bayes Models
3.4.3.4 Make Predictions with a Naive Bayes Model
3.4.3.5 Na?ve Bayes Algorithm Advantages and Disadvantages
3.4.4 K-Nearest Neighbors– Classification
3.4.4.1 Algorithm
3.4.4.2 K-NN Algorithm Advantages and Disadvantages
3.4.5 Random Forest
3.4.5.1 How does The Random Forests Algorithm work?
3.4.5.2 Feature Importance
3.4.5.3 Random Forest Algorithm Advantages and Disadvantages
Chapter 4 Crash Data Collection and Data Description
4.1 Introduction
4.1.1 Hong Kong Transportation Department Accident Data
4.1.2 Variables Considered In the Study
4.2 Data Preparation
4.2.1 Based on Accident
4.2.2 Based on Vehicle
4.2.3 Based on casualty
4.3 Data Pre-Processing
4.3.1 Missing Data Treatment
4.3.2 Data Normalization
4.4 Estimation of Accuracy in Classification
4.5 Models Selection by Performance Evaluation
Chapter 5 Data Analysis and Modeling Results
5.1 Statistical Models Results
5.2 Machine Learning Models Analysis Results
5.3 Experiments and Results
5.3.1 CART Experimental Results
5.3.2 Support Vector Machine Results
5.3.3 K-Nearest Neighbor Results
5.3.4 Gaussian Na?ve Bayes Results
5.3.5 Random Forest Results
5.4 Results Comparison of Machine Learning Models
5.5 Summary
Chapter 6 Discussion of Findings
6.1 Sensitivity Analysis
6.2 Comparison of Variable Impact on Crash Severity from ML Models
6.3 Summary
Chapter 7 Conclusion and Recommendations
7.1 Conclusion
7.2 Recommendations
References
Acknowledgements
List of Figures
List of Tables
List of Acronyms
本文編號(hào)：3974965
本文鏈接：http://www.wukwdryxk.cn/kejilunwen/daoluqiaoliang/3974965.html
最近更新
教材專著
