Analyzing median gap crossing patterns using negative binomial regression model

Median gap crossings are required for guaranteeing the mobility as well as safety of urban traffic. Uncontrolled vehicular and pedestrian crossings at median gaps lead to traffic jams and accidents; hence, accurate behavioral modelling becomes crucial. The traditional models are afflicted by their inability to deal with overdispersion in crossing data suitably, which leads to inadequate predictions. In the current study, a Negative Binomial Regression (NBR) model is employed to analyze pedestrian behavior in crossing an urban road section from Senneer Kuppam to Avadi Check Post. Through an examination of a data set of 100,000 crossings spread across 13 months with the aid of CCTV as well as observational study, peak hours of crossing, the impact of vehicular traffic, and location-specific variations were determined. The NBR model is more effective compared to Poisson regression with its smaller Akaike Information Criterion (AIC) values and higher log-likelihood values. Also, the study highlights the significance of pedestrian infrastructure development and AI-based traffic monitoring. Future predictive studies based on traffic control must include deep learning models combined with real-time observation methods.