Relationship between Highway Geometric Characteristics and Accident Risk: A Multilayer Perceptron Model (MLP) Approach

The traffic safety of mountain highway has always been one the taking point. This study aims to collect road design data in large-scale research and analyzes accident risk geometric alignment. Accordingly, a method based on satellite maps clustering algorithms is proposed calculate alignment plane its longitudinal section. reliability was verified Nanfu Chongqing, China. planar sectional geometries four highways Chongqing were obtained by above method, corresponding 36,439 accidents which occurred from 2010 2016 used as objects. geometry analyzed SHAP MLP theories. results show that fitting prediction abilities model are better than those negative binomial model, correlation coefficient improved 33.2%. In addition, compared with can estimate more accurately flexibly complex nonlinear relationship between independent dependent variables.