Abstract:
Pedestrians are the most vulnerable group of road users. In order to protect the pedestrians from
fatalities while crossing through at grade crosswalks, grade separated crossing facilities are
provided. But they are sparsely used by the pedestrians. Using zebra crossing has become part
of their habitual action. Pedestrians are not willing to use footbridge even though they are safe.
The study aims to identify factors influencing the non-usage of footbridge using perception
survey. Machine learning techniques was used to analyze the data. Different algorithms such
as random forest, decision tree, support vector machine, k-nearest neighbor and logistic
regression were compared to get best model. Random forest outperformed all other model with
an accuracy of 83%. The weightage of factors was computed using random forest. It was found
that frequency of using footbridge has a high weightage among other factors. It was followed
by easiness, stressful, weather, tiresome, night, bad infrastructure, heavy traffic, height fear,
unfamiliar location, accessibility, hurry, good illumination, education, age, pedestrian accident
history and occupation. Some counter measures such as escalator, alerting poster, fine for
illegal crossing, seating arrangement, security staff are suggested in the study to identify the
change in usage of footbridge among non-users. Most of the pedestrians are willing to use
footbridge if escalator or elevator is provided. To promote footbridge utilization, a usercentered strategy is necessary.
