Risk Estimation of Signalized Intersection Accident

I. INTRODUCTION

In urban area most of the traffic accident happened at intersection. And analyzing intersection accident by their types is very important because it is proved that total accident at intersection have very poor prediction.

II. RESEARCH BACKGROUND

After motorization started at the beginning of this century, Traffic Accidents became a heavy financial lumber on society. After that various measures were adopted to improve the safety condition. In Japan, although many countermeasures have been, traffic accidents are still mounting. Effective countermeasures against intersection accidents are immediately predicted my the Negative Binomial Model. These distinctive features are assimilated in the model to articulate accident model accuracy.

III. RESEARCH RESULTS

For Right-turn Accident (AG1) twelve factors for Po and eight factors for Pf. Likelihood ratio index is 0.51 and average probability for Po and Pf are 0.133 and 1.28^10-6 respectively.

For Right-Angle Accident (AG2) fifteen factors for Po and nine factors for Pf. Likelihood ratio index is 0.41 and average probability for Po and Pf are 0.159 and 5.25^10-6 respectively.

For Left-turn Accident type ten factors for Po and eight factors for Pf. Likelihood ratio index is 0.47 and average probability for Po and Pf are 0.295 and 3.8^10-6 respectively.

The spatial analysis and maps developed in GIS illustrates the relative probability of each type of accident risks of collision occurrence resulting from Accident Modeling. It is suggested that this methodology could easily maintained with periodic updates of data and accident information and hence giving possible and important factors or reason responsible for that type of accident and their corresponding countermeasures and thus creating a Dynamic Model from which traffic accident in intersection in urban area can be easily monitor and takes appropriate countermeasures for each type of accident. Lastly some Hazardous Intersections. are identified comparing modeling results and Rate Quality Control Method.

Some intersections of different prefectures are Remarkably similar with respect to flow, speed, road environment and traffic characteristics irrespective to their prefecture. This particular observation stimulate to apply the statistical technique of "Discriminant Analysis". This is because different region have diverse accident characteristics and only one microscopic model can't be applied in all region. To achieve the aforementioned methods, similar groups were determined based on the geometric criteria at the outset and then "Discriminant Analysis" were applied to find out the best discriminating variables between the groups and also to dwindle over dispersion (i.e. the variance of accident frequency data is greater than its mean) in accident database. Statistical results conferred an excellent Model for the location of different groups and also means of using the model under different "Confidence Level" assumptions.