Development Of Pedestrian Safety Index Model At Un- Signalized Intersection

Abstract

Pedestrian safety has become a critical issue in developing countries where rapid urbanization
and motorization have outpaced the development of pedestrian-friendly infrastructure. Inadequate crossings,
limited enforcement, and insufficient pedestrian facilities have increased the frequency of pedestrian–vehicle
conflicts, particularly at unsignalized urban intersections. This study focuses on developing a Pedestrian
Safety Index (PSI) to quantify pedestrian safety levels by considering human, environmental, and traffic-
related parameters. Data were collected through videography over three-hour observation periods, covering
the movement of 2,356 pedestrians. Variables such as age, gender, crossing speed, mobile phone usage,
vehicle and pedestrian volumes, police presence, road width, number of lanes, and sidewalk conditions were
analyzed. Using Multiple Linear Regression (MLR), the model achieved an R² value of 0.753, indicating
that 75.3 percentage of the variation in pedestrian safety could be explained by the selected parameters. To
further enhance predictive accuracy, an Artificial Neural Network (ANN) model was developed, producing
Sum of Squares Error (SSE) values of 447.64 (training) and 221.53 (testing), and relative error values close
to unity, demonstrating high model reliability and generalization. The study concludes that pedestrian safety
is significantly influenced by both behavioral and roadway factors. Integrating these parameters through
statistical and machine learning models offers a reliable framework for assessing and improving pedestrian
safety. The developed PSI serves as a valuable decision-support tool for urban planners and traffic engineers
to design safer, more inclusive intersections and to prioritize safety interventions in rapidly developing urban
environments.