Accident analysis of long and steep longitudinal highway slopes based on multi-source data under the background of Internet of Things

Abstract

Under the background of "Internet of Things + big data", the accident analysis and safety prevention of long and steep sections of highway have attracted the attention of academic circles. Reasonable analysis of safety influencing factors can contribute to ensuring traffic safety of highway projects and improve prevention technology. Therefore, based on multi-day and multi-source traffic accident data in Guizhou province, this paper constructs a Multinomial logit(MNL) model considering multiple accident causes. Firstly, data related to accident causes are screened out from multi-source data, and these data are divided into two categories of traffic accident characteristics and road environment characteristics. Secondly, based on the multi-day survey data, the MNL model was established by using python to conduct regression analysis on the relationship between the accident influencing factors and the accident severity under the general road section and the long and steep road section respectively, and the two were analyzed and compared based on the significance factors. Finally, the accident data of long and steep longitudinal slope sections are selected to do univariate regression analysis of the annual accident number and each influencing factor. An empirical experiment was carried out in Guizhou Province. The results show that the accident type, vehicle type and accident location are the obvious differences between the long and steep slope road accidents and the general road accidents. The coupling effect of slope length and slope, traffic flow and truck proportion have great influence on the number of annual accidents. This study can provide theoretical support for the improvement of highway accident prevention technology in mountainous areas.