Maximum Likelihood Estimation - Negative Binomial Regression Traffic Conflict Prediction Model Considering Variable Interactivity

Using video recognition technology to extract traffic conflict and traffic flow data of 121h of Chengdu Hongguang Road (including 6 intersections), select the initial variables of the model through qualitative analysis, and consider the interaction between different variables to add a new interaction variable. Finally, a negative binomial distribution traffic conflict prediction model is constructed. We use the AIC criterion to compare the accuracy and fit of the old and new models and use the maximum likelihood estimation method to estimate the newly established model parameters. The results show that the maximum likelihood estimation -negative binomial regression prediction model considering of the interaction variable can significantly improve the prediction accuracy of traffic conflict times.