標題: 大台北地區小客車肇事影響因素之研究
A Study on Potential Factors influencing Passenger Car Accident in Taipei Metropolitan Area
作者: 蕭育裴
Hsiao, Yu-Pei
黃承傳
Cherng-Chwan Hwang
運輸與物流管理學系
關鍵字: 小客車;計程車;肇事;卜瓦松迴歸;負二項迴歸;零膨脹;passenger car、、、、、zero-inflated;taxi;accident;poisson regression;negative regression;zero-inflated
公開日期: 2010
摘要: 小客車是一種快速、便利的運輸工具,符合現代人快速步調之需求,然而小客車也是很容易發生肇事的車種,在交通量龐大的大台北地區更是如此,小客車肇事的原因往往會歸咎於駕駛的人為疏失,包括違規及各種失誤,若是駕駛人具高危險特性,乘客之生命安全即受到威脅。目前國內對於肇事次數的研究多著重在公路特性的影響,很少有「駕駛人特性」之肇事次數研究。 本研究的宗旨是運用卜瓦松迴歸模型的架構,試圖建構自用小客車肇事模式,探討駕駛的各項特性(包括社經變數、違規情況、車輛情況)與其發生肇事次數之間的關連性;由於肇事次數資料因為零值過多(發生0次肇事樣本占多數),而呈現過度離散(over-dispersion)及零膨脹(zero-inflated)的疑慮,因此除了使用卜瓦松迴歸模式以外,建構的過程也加入負二項迴歸、零膨脹卜瓦松迴歸及零膨脹負二項迴歸模型。另外,為了探討一般駕駛與職業駕駛的肇事模式是否不一樣,還建構計程車駕駛的肇事次數模式,進而比較職業駕駛與一般駕駛在駕駛特性上的異同。 研究結果發現自小客車較適合零膨脹卜瓦松模式,影響自小客車肇事次數的變數有「使用測速雷達」、「使用行動電視」、「車輛總行駛公里數」、「超速」、「超車、違規轉彎」、「駕駛為男性」、「駕駛60歲以上」;計程車則適合零膨脹負二項模式,影響計程車肇事次數的變數有「每日行駛公里數」、「每日營運時數」、「每日載客趟數」、「超速」、「違規超車、轉彎」。其中超速、違規超車或轉彎兩種違規都顯著影響自小客車及計程車肇事。
Passenger car is a rapid and convenient vehicle and it is accorded with the demands of the modern people. However, passenger car could be a high-risk motor vehicle especially in heavy-traffic Taipei metropolitan area. Driver’s behavior, including error or to drive against traffic regulations, play an important role in traffic accident occurrence. If a driver has higher risk to cause crash, it will threaten the life safety of passengers. Most past studies that related to the number of accident put their focus on the highway factors. Very little research has been done to explore the driver’s characteristic factor in relation to motor vehicle crashes. The primary objective of this research was to develop the Poisson regression model that explains the relationship between drivers’ character factors and car accidents. When the number of observed zeros exceeds the expected number of zeros under the Poisson assumption, zero-inflated and over-dispersion occurs. If this phenomenon is not taken into account, the traditional modeling approach used to analyze this type of zero-inflated data has the potential to result in biased model coefficients. Hence, a number of approaches (such as the negative binomial model and the zero-inflated model) are available in the literature to deal with this problem. In order to analyze the different between passenger drivers’ characteristic and professional drivers’, this research also develop taxi driver mode. Passenger drivers were compared to taxi drivers within the same time period of this study. The estimated results in passenger car indicate that ZIP model has better performance. The results showed that “Using radar detector”, ”Using mobile television”, ”Total mileage number”, ”Speeding violation”, ”Illegal passing or turning”, ”Driver who is male”and”60 and older year-age-group” increase the likelihood for accident involvement of passenger car. Moreover, ZINB model has best consequence on taxi car. The results showed that, ”Speeding violation” and ”Illegal passing or turning” increase the likelihood for accident involvement of taxi. Both model illustrated the significance of “Speeding violation” and “Illegal passing or turning”, on the frequency of accident occurrence.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079836519
http://hdl.handle.net/11536/47991
Appears in Collections:Thesis