The Impact of Pedestrian Crossing at Unsignalized Intersection Using Machine Learning, Binary Logistic Regression, and NARX

Abstract

Fajaruddin Mustakim*, Azlan Abdul Aziz, Mohammad Nazir Ahmad, Riza Sulaiman, Rabiah Abdul Kadir, Othman Che Puan and Muhammad Nizam Zakaria

In 2024, Malaysia experienced a total of 532,125 road accidents and 5,364 fatalities. Meanwhile pedestrians recorded more than 550 casualties each year and consistently placed third rank after motorcyclist and passenger car. This study aims to analyze the influence of pedestrian crossing at selected unsignalized intersections during the vehicle’s manoeuvres. The eleven selected sites were based on blackspot location and the study focused on comparison between intersections with and without pedestrian bridge facility. In the early stage the study determined the pedestrian crossing characteristic throughout the day and the observations were concentrated on three peak hours which is the morning, midday, and afternoon. Next determine the frequency of traffic patterns involving pedestrian crossing (PC), traffic conflict (TC) and motorcycle crossing (MC). Traffic characteristic fluctuation analysis based on PC, TC, MC, and approach speed (AS) were carried out. Finally, this study manages to develop right-turn motor vehicles (RMV) considering pedestrian crossing and other traffic variables by adopting Binary Logistic Regression (BLR), Machine Learning base on Neural Net Fitting and Nonlinear Autoregressive Exogenous model (NARX). The RMV model’s calibration using 838 datasets and involving eight predictors that influence the vehicle's manoeuvres. The study reveals that pedestrian crossing, traffic conflict and traffic volume affect the RMV to accept shorter gap and providing the pedestrian bridge has a positive impact for vehicle manoeuvres. In addition, the result from machine learning using scale conjugate gradient algorithm achieved mean square error within 10% or performed accuracy 90%, in the RMV model. Although the pedestrian bridge has been provided at the intersection, in a few cases the pedestrian refuses to benefit the facilities. This study recommends the implementation of autonomous vehicles (AV) and electric vehicles (EV) that are equipped with internet of things (IoV) and vehicle-to- everything communication (V2X) as part or partial solution in reducing traffic accidents.

HTML PDF