Leveraging Deep Learning for Pedestrian and Cycle Flow Prediction in Urban Environments

Benitez Martinez, Carlos Manuel (2025) Leveraging Deep Learning for Pedestrian and Cycle Flow Prediction in Urban Environments. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

The accurate prediction of urban mobility is essential for developing sustainable and responsive smart cities. Monitoring and predicting urban mobility remain a challenge. However, previous work integrates spatiotemporal data as static variables, and spatial data available is limited. This study focuses on predicting pedestrian and cyclist flows in Dublin city on an hourly basis with the integration of data from multiple sources, including data from footfall (13 locations) and cyclists (6 locations), with three different features as input, dynamically varying time (12 features), weather (12 conditions) and spatial (2 features). This study proposes the usage of advanced deep learning architectures, Long Short-Term Memory (LSTM) and hybrid Convolutional Neural Network combined with Long Short-Term Memory (CNN-LSTM) models, with capabilities to capture complex spatiotemporal information and understand temporal context in urban mobility patterns. By comparing these models against traditional machine learning baselines such as Random Forest and XGBoost, the study demonstrates superior performance of deep learning in handling volatile pedestrian and cyclist flows. The work indicates that a scalable pipeline can improve open-source cloud resources to achieve reproducibility in data-driven urban planning. The findings noted how integrating dynamic weather data and temporal features, excluding spatial features shows significant improvements in error reduction of more than 50% for pedestrians and 30% for cyclist flows. The results show that LSTM and hybrid CNN-LSTM model without spatial features outperform Random Forest and XGBoost models according to three metrics commonly used in regression: mean absolute error, root mean squared error and coefficient of determination.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Kelly, John UNSPECIFIED
Uncontrolled Keywords:	Heterogeneous data integration; pedestrian prediction; cycle flow prediction; deep learning; CNN; LSTM
Subjects:	Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning H Social Sciences > HE Transportation and Communications > Urban Transportation
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	30 Jun 2026 16:59
Last Modified:	30 Jun 2026 16:59
URI:	https://norma.ncirl.ie/id/eprint/9410

Actions (login required)

View Item