Abstract:
The transportation facilities are used for the efficient transfer of goods and services. Even
though the safety is the major concern for a transportation engineer, there are many conflicts
occurs creating severity to pedestrians and vehicular users due to lack of proper safety
measures. Among the road users the pedestrians are the ones more vulnerable to conflicts.
When conflicts occur, pedestrians came to direct impact without any protection rather than
other road users such as vehicle users and passengers having some means of protection. Many
studies have been done related to safety of pedestrians near intersections where the violation
behaviour is more. But there is lack of study in midblock locations which was identified as the
gap for study. The study mainly aimed to analyse the perception of pedestrians towards bus
stop, crosswalk and sidewalk and to find out the violation behaviours while using these
facilities followed by model creation to find out the safety of pedestrians near urban bus stops.
The data collected through questionnaire and video were used for analysis of perception of
pedestrians towards each facility, violation behaviour and model creation. The modelling was
done with 2 machine learning techniques Regression and Classification using various
algorithms at initial stage to find out the best model with more performance. The attributes
selected for the model creation for predicting PET (Post Encroachment Time) includes the
perception data, behavioural data and vehicular data with a total count of 33 variables. The
initial stage analysis done by separating 4 regression variables and 29 classification variables.
The analysis indicates 9 variables are influencing PET, based on which best machine learning
models were created using Random Forest giving R2 Value of 0.84 and MSE of 0.55 for
regression and Accuracy of 0.82 and better F1 Score for classification. Also, the weightage of
factors was derived using the best model and used for development of Safety Index, which
indicates the safety level of bus stop with help of K-means Clustering in SPSS. Also, the
required recommendations for improving the safety level of bus stops were given for selected
10 bus stops for the study
