Identifying the principal factors influencing traffic safety on interstate highways

Abstract

This study aims at identifying the principal factors influencing fatal, nonfatal injury and non-injury traffic crashes on urban and rural interstate highway segments using a statistical approach called principal component analysis. Initially, fourteen explanatory variables including segment length, annual average daily traffic (AADT), weekday/weekend, hour of the day, urban-rural designation of the segment, median type, pavement surface condition, roadway geometric characteristics, weather, number of lanes, and drivers' age and gender, and the accident year were considered. Separate principal component analyses are performed for the three crash categories and the aggregate dataset. The results of the analyses show that, regardless of the crash categories used, seven principal components accounting for over 70% of the variances in the original datasets were retained. In addition to the overall dimensional reduction of the original dataset by 50%, the results suggest that the key variables contributing to the crashes across the categories of the accident types observed on the freeway segments are insignificant. The retained principal component loadings of the factors PC1, PC2, PC3, and PC5 revealed the fact that the number of lanes, the median type, and the AADT of the segments are highly correlated and represented by the first factor (PC1). Similarly, other interrelated factors such as the prevailing weather and the pavement surface condition (wet, dry, snow), the hour of the day and the lighting condition, the drivers' age and gender are well represented by PC2, PC3, and PC5 respectively.

Choice behavior of commuters' rail transit mode during the COVID-19 pandemic based on logistic model

“…(2019) used principal component analysis and regression analysis to study the relationship between the choice behavior of electric vehicles and the performance characteristics of electric vehicles, users' perception of the benefits of electric vehicles and relevant policies. Kassu and Hasan (2019) used principal component analysis and regression analysis to study the influencing factors of fatal injurious, non-fatal injurious and non-injurious traffic accidents on the road. Al-Ghamdi (2002) used regression analysis method to study the impact of accident location, time, cause, vehicle type, age of the perpetrator and other factors on the severity of the accident.…”

Traffic Crash Severity Prediction—A Synergy by Hybrid Principal Component Analysis and Machine Learning Models

“…Based on the results of the previous study, the author of this study tried to enhance the prediction accuracy of RTC severity by investigating the effect of dimension reduction (by using PCA) of a crash dataset on the performance of NN and SVM models in RTC severity prediction. PCA was used by scholars to find out the important features affecting crash severity [28][29][30].…”

Hybrid feature selection-based machine learning Classification system for the prediction of injury severity in single and multiple-vehicle accidents

“…To overcome this limitation, a number of researchers have reported the utility of feature selection algorithms such as principal component analysis (PCA)-based ANNs and PCA-based MLR in modeling the retention times of a variety of volatile organic compounds [ 27 ], ground-level ozone and the factors that influence its concentrations [ 28 ], and internal glasshouse humidity in North China during the winter [ 29 ], evaluation of effect of E-beam irradiation on ready-to-eat food [ 30 ], rain water quality modeling [ 31 ] and development of pistachio sorting system [ 32 ]. In the traffic and transportation domains, feature selection algorithms have been used to analyze mode choice [ 33 ], identify hotspots on roads [ 34 ], and investigate key factors affecting injury severity on rural and urban highway segments [ 35 ].…”