Abstract

Public transport systems increasingly rely on data-driven approaches to understand passenger demand and improve operational efficiency. However, conventional bus stop classification methods are typically static and fail to capture variations in passenger accumulation patterns over time. As a result, certain bus stops may experience high levels of congestion without being identified as critical nodes within the network. This study addresses this gap by developing a data-driven framework to model passenger accumulation and identify hidden demand patterns at bus stops using Automated Fare Collection (AFC) data. The research investigates whether Regular Bus Stops exhibit homogeneous demand behavior and to what extent hidden high demand stops can be identified through machine learning and behavioral analysis. The study is based on a large-scale AFC dataset capturing passenger boarding activity over multiple weeks. A hybrid analytical approach is adopted, combining supervised classification and unsupervised clustering techniques based on passenger accumulation, while K-Means clustering is applied to segment bus stops into distinct behavioral types based on temporal demand patterns, including peak-hour and day-of-week variations. The findings demonstrate that Regular Stops are not homogeneous and can be categorized into multiple behavioral clusters, including commuter-driven, weekend-oriented, and high-demand backbone stops. The results further show that hidden congestion is not randomly distributed but is strongly associated with specific behavioral patterns. Weekend oriented stops exhibit a disproportionately high likelihood of hidden demand despite not being classified as major transport nodes. Additionally, high-demand evening stops display extended periods of elevated accumulation beyond peak hours, indicating the presence of hidden congestion not captured by traditional peak-based analysis. The study concludes that passenger demand at bus stops is influenced not only by volume but also by temporal distribution and behavioral patterns. The proposed framework provides a practical approach for identifying overlooked high-demand locations and supports more effective transport planning. It is recommended that transport authorities incorporate temporal demand segmentation into stop classification and adopt data-driven methods for service planning, particularly for weekend and event-driven demand. Future research may extend this framework by integrating spatial analysis, applying the methodology to other transport modes, and exploring real-time applications using streaming data.

Publication Date

5-19-2026

Document Type

Thesis

Student Type

Graduate

Degree Name

Professional Studies (MS)

Department, Program, or Center

Graduate Programs & Research

Advisor

Ehsan Warriach

Comments

This thesis has been embargoed. The full-text will be available on or around 12/16/2026.

Campus

RIT Dubai

Download

Share

COinS