| dc.description.abstract | Accident analysis done by traffic–related departments has been continuously discussed and followed. The reason for this is to confirm the causes of accidents and further suggest improvement strategies in order to lower accident rates or alleviate the severity of accidents. Related departments have figured out the number of accidents that occurred on different highway sections over the past 10 years, setting up road signs on prone traffic accidents so as to raise driver awareness and lower the number of accidents. However, these signs can only provide statistical information to drivers and are unable to be updated along with time, traffic and weather conditions. Therefore, this research is aimed at providing dynamic information of prone traffic accidents for drivers as a reference.
The research focused on accidents that took place at the northern section of Taiwan highway number one and performed research on relevant statistics, risk analysis and potential analysis. Firstly, the road section in question was divided up based on linear geometric characteristics and a Zero-inflated Negative Binomial (ZINB) Regression Model was used to establish a Model of accident rates. After that, the accident rates were used to predict simulation results, which was then followed by a risk analysis to obtain values at risk and utilized potential analysis to categorize different values of risk. At the end, geospatial analysis was used to produce a potential map. Research outcome had revealed that the four significant variables affecting accident rates of the northern section of highway number one are, vertical grade, traffic flow, V/C value, and exposure. Lastly, analysis on data recorded on 2nd April, 2015 was performed and it was realized that the number of high accident potential road sections surged to a maximum at 7 am in the morning, while the, Jubei interchange stayed at high to extremely high potential from 6 am in the morning to 12 am in the evening. | en_US |