|
Ajay Singh Meena
Deepak Mathur
Keywords:
Jaywalking, Pedestrian behavior, Pedestrian safety, Statistical analysis, ROC curve, Machine learning models, Support vector machine (SVM), Crossing speed.
Abstract:
People frequently jaywalk and engage in uneven or illegal crossing at signalized crossings in developing countries, which significantly increases the likelihood of deadly accidents. Consequently, the level of service quality at signalized crosswalks diminishes. To examine and simulate pedestrian jaywalking behavior at major signalized crossings in an urban Indian city, an observational and field study is conducted. Pedestrian flows, geometric features, and crosswalk characteristic were collected for this study using a video-graphic technique. Multiple Correlation and exploratory factor analysis were then employed for statistical analysis.
According to the findings, there are seven main parameters that affect the pedestrian jaywalking index: flow physiognomies, dimensions, road features, arrival attributes, crossing patterns, and physical attributes. With a 89.40% success rate, a binary logit model identified seven key variables that influence the likelihood of pedestrian jaywalking: gender, the number of lanes, the width of the crosswalk, the crossing pattern, the type of signal upon arrival, the existence of guardrails, and the average pedestrian delay. An outstanding degree of discrimination is represented by the ROC curve’s (0.892) area under the curve, which helps improve pedestrian safety.
The study focused on pedestrian flow parameters including crossing speed and waiting time, looking at pedestrian variables (age, gender, baggage, and tread pattern) on crossing patterns.
A range of machine learning models were trained and assessed, like SVM, multilayer perceptron’s, decision trees, & Bayesian techniques. When compared to other models, the SVM model showed the highest precision in forecasting the likelihood and velocities of pedestrian crossings.
|
|
International Journal of Recent Research and Review
ISSN: 2277-8322
Vol. XVII, Issue 2
June 2024
|
PDF View
PUBLISHED
June 2024
ISSUE
Vol. XVII, Issue 2
SECTION
Articles
|
