A Study on Freeway Traffic Accident Using Survival Analysis Method

• Main Authors: Chia-Ying Chou, 周佳穎
• Other Authors: Jiann-sheng Wu
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/rpgb5x

碩士 === 國立中央大學 === 土木工程學系 === 106 === With the progress of society, the domestic traffic network has gradually improved. The National Freeway No. 1 connecting the north and the south has become the main artery of the Western Corridor of Taiwan. Because the tandem area is dominated by the metropolitan area, the traffic volume is also more intensive. The road environment of the expressway is relatively closed and has the characteristics of large traffic flow and fast speed. It will inevitably lead to traffic accidents and back-blockages, resulting in negative impacts on driving safety, time lapse and financial losses, as well as higher social cost. Therefore, this study conducts follow-up analysis on National Freeway No. 1 with a relatively high traffic volume and number of accidents. It hopes to identify the cause of the accident and then propose improvements to reduce the accident rate or reduce the severity of the accident. In this study, the factors that may affect the accident are divided into 12 variables of geometric characteristics, traffic flow characteristics and climate, and reference the design specification or define the impact variables through analysis and definition to construct the Cox proportional hazards model through parameter review, model matching test and equal proportion test to find out significant variables. The relative risk and the number of accidents in the road section are estimated by the significant variables. The research validated that 25% of the accidents interval would be more than 503 days, 50% of the accidents interval would be more than 203 days and 75% of the accidents interval would be more than 62 days. After verification, the “slope”, “flow”, “exposed quantity” and “temperature” four variables are significant, indicating that the above variables can effectively explain the interval of accidents.