Utilizing Latent Class Logit Model to Predict Crash Risk

ABSTRACT:Real time traffic crash risk prediction is very important for proactive safety management systems. In this paper, a new model is developed to predict real time traffic crash risk based on some elaborately selected characteristics of traffic flow. Compared with the dominant conventional matched case-control logistic regression model which is based on the assumption that all the traffic crashes share the same contributing factors and ignores the variability of traffic flow and heterogeneity of crash causations, the latent class logit model can considerably account for the unobserved heterogeneity. Data from Shanghai urban expressway between April to June 2014 are utilized to build the latent class logit model. The collected data are divided into a training set for model building and a validation set for model evaluation by random selection. Some criteria including Akaike information criterion (AIC), area under receiver operating characteristic curve (AUC), sensitivity and specificity are used to evaluate the model quality. The experimental results show that the overall accuracy of the latent class logit model is 70.23% and AUC of the model is 0.7576, which are both higher comparing to the traditional widely used case-control logistic regression model.

Zhiqiang Huang, Zhen Gao*, Rongjie Yu*, Xuesong Wang, Kui Yang. Utilizing Latent Class Logit Model to Predict Crash Risk, 16th IEEE/ACIS International Conference on Computer and Information Science (ICIS 2017), Wuhan, P.R. China, 2017.5.24-26. (EI)