| 摘要: | 交通事故發生在現今各國社會層面中是項重要的問題，而運輸安全問題長久以來也為我國所面臨的主要難題之一。依交通部道路交通安全督導委員會2017年公布數據顯示，台灣地區每年平均有3,000人左右死於交通事故，其中又以機車交通事故件數占比最高且當事人之死傷程度最為嚴重。因此，如何改善機車善運輸安全以降低交通事故的嚴重性已為當務之急。本論文研究對象主要針對在桃園地區發生機車事故之第一當事人為觀測對象，資料來源為內政部警政署民國103到105年桃園市交通事故中A1、A2類交通事故調查報告表相關統計資料，以統計模型如羅吉特迴歸以及多元羅吉特迴歸模型進行實證研究分析，而模型中將機車道路事故影響變數分類為當事人特徵、環境因素、道路型態以及駕駛主要肇事因素等。
由羅吉特迴歸模型分析實證結果顯示，以交通事故機車駕駛是否涉入重傷或死亡事故為衡量指標時，無照駕駛、酒駕、夜間、圓環與駕駛主要肇事因素等皆導致駕駛涉入重傷或死亡事故的機率提高；而由多元羅吉特迴歸模型分析實證結果顯示，以交通事故機車駕駛涉入之傷亡程度為衡量指標時，學生、從業人員、酒駕、夜間、圓環與駕駛主要肇事因素等皆導致駕駛之傷亡情形更為嚴重。另外，以交通事故機車駕駛是否違反主要肇事因素為衡量指標時，分析結果顯示學生與從業人員等駕駛較容易有肇事行為。;Traffic accidents are an important issue of modern society, and transportation safety has been a key concern in Taiwan for a long while. According to 2017 statistics from the Ministry of Transportation and Communications, an average of 3,000 people died from traffic accidents in Taiwan annually. Motorcycle accidents correspond to the highest fatality rate and most serious degree of injury. Therefore, improving the safety of motorcycle transportation and reducing its casualties has become a priority. Using data released by the National Police Agency of Taoyuan City’s A1- and A2-type traffic accidents from 2014 to 2016, this study aimed at the person who was injured or died in the motorcycle accident. Using logit and multinomial logit model regressions, we empirically analyzed motorcycle accidents to determine characteristics of the person, environmental factors, road types, and major causes of accidents cause motorcycle fatalities.
Results of the logit regression model show that that several factors significantly increase the fatality rate of motorcycle traffic accidents, including driving without a license, drunk driving, nighttime driving, roundabouts, and major causes of accidents cause motorcycle fatalities. Results of the multinomial logit regression model show that students, employees, drunk driving, nighttime driving, roundabouts, and major causes of accidents cause motorcycle fatalities will lead to more serious casualties in motorcycle accidents. Additionally, using another multinomial logit regression model and assigning major causes of accidents cause motorcycle fatalities as dependent variables in model, we found that drivers who are students or employees are more likely to have accident-prone personalities and therefore be involved in motorcycle traffic accidents. |
