| Summary: | 碩士 === 逢甲大學 === 交通工程與管理研究所 === 89 === As a result of the situation that some signals or turning regions can’t correspond to the demand of passing vehicles entirely, or drivers desire temporary convenience, then they will drive violatively, that will lead to conflict. This study strives for prescription of modifiable question, so only collects dynamic, short-term, and plentiful data of traffic violations and conflicts. This study uses violative data as exposure of conflicting events. And it collects person, vehicle, road, and environmental data for four signalized intersections in Taichung City. In dealing with different relevant questions, this study uses several methods such as Cox’s PH model, accelerated failure time model, logistic model, double layers logistic model, discriminate function, or dynamic model for inference, then finds the consistence of these methods, and looks for causes of questions. This study analyzes and compares the number of violation and conflict with suitable models between Taichung City and Taichung County. The results show that, compared with Monday, Sunday decreases the probability of occurring earlier either in four seasons or at four intersections. For four seasonal models, left turning makes conflicting probability increasing, motorcycle makes probability decreasing. For four related intersections models, winter makes conflicting probability decreasing except the intersection of Chung-Kang road and Kung-San road. This research applies double layers logistic model either in seasons or at intersections, only if the influence of inclusive value for some model is remarkable, then goodness-of-fit index of this model will be better than that of logistic model. If the frequency of violations and conflicts are classified into more categories in discriminate analysis, the correctly discriminating rate decreases. By applying cluster analysis to classification, the correctly discriminating rate will be averagely 7% higher than that by the evenly classified method. This research only focuses on application tests of models, so the results aren’t suitable for direct advice about traffic management in Taichung City. The reason is that people don’t participate in four seasonal surveys continuously, and they have different viewpoints on the definition of conflicts.
|
|---|
