Roadside Safety Assessment Of The Low-grade Highway In Mountainous Areas Based On Projection Pursuit

Low grade highway in mountainous areas always walks along the mountain trend with steep slope and sharp turn, which limited to the difficult terrain. Besides, it is easy for the vehicle to rollover or impact fixtures and so on roadside accidents.as the roadside environment is dangerous, once there is a roadside accident always characterized with group death and group injury. Therefore, quantifying the roadside hazard and strengthening the roadside protection measures is of momentous significant to improve traffic safety of the low-grade road in mountains.According to the road alignment materials and roadside accident statistics in recent five years, combined with the characteristics of low-grade highway in mountainous, analyzes the roadside accidents characteristics systematically. Based on roadside accident characteristics analysis, studies the effect of road geometry, traffic flow characteristics and roadside environment on the roadside hazard.Secondly, we introduce the common evaluation methods and applicability, advantage and disadvantage. however,those method have clear limitations, especially regarding weights calculated subjectively, evaluation level thresholds without evidence and cannot be tested. So aiming at these problems, an exploratory analysis method based on data-driven. which is roadside safety assessment based on projection pursuit method is proposed. And then, multi-agent genetic algorithm is used to solve the model. In order to determine the rating classification number and threshold level of the roadside hazard, Fisher optimal partition method is applied to division of the projection value sequence.Finally, based on the principle of selecting evaluation indexes, they constructs the roadside safety evaluation system of the low-grade highway in mountainous based on identification of impact factors. Besides, in order to verify the validity and rationality of the model, the low-grade highway in Chengde city is selected as a case study by using the proposed model.This paper applies projection pursuit technology to reflect the relationship between evaluation index and roadside hazard. And uses the Fisher optimal partition method determine the classification number and the corresponding threshold level. The result shows that the method is feasible, reasonable. Most important, the proposed model have verifiability and potential replication.