MRes Project
Abstract Link to heading
This dissertation presents a novel investigation into the integration and impact of data quality on predictive modeling for urban pedestrian flows. Leveraging a robust analytical framework, it explores the challenges and opportunities presented by urban observatory datasets in forecasting pedestrian movements within city environments. Through meticulous analysis, employing both linear and LSTM (Long Short-Term Memory) models across various states, the study delves into the nuances of anomaly prediction and the critical role of data quality in enhancing model accuracy. Key findings illustrate the limitations of univariate models in the context of pedestrian data, pinpointing the necessity for incorporating periodicity features and addressing data quality issues to improve predictions. The research advances the field of urban data analytics by demonstrating the potential of artificial intelligence to refine situational awareness tools, ultimately facilitating more informed urban planning and management decisions. This work not only contributes to academic discourse on urban informatics but also offers practical insights for policymakers and urban planners aiming to leverage data-driven approaches for city development and pedestrian safety.