Data Mining Applications for Pedestrian Behaviour Patterns at Unsignalized Crossings

This study analyses pedestrian behaviour patterns at unsignalized crossings by using multiple Data-mining approaches, aiming to improve pedestrian safety by understanding the relationship between movement patterns, location, and infrastructure. Utilizing the STATS19 dataset from the UK Department for Transport, applied data analysis techniques, including heatmap visualization, association rule learning, and Principal Component Analysis (PCA) with clustering, to identify high-risk behaviours and provide targeted interventions. Heatmap visualization identifies spatial patterns and high-risk areas, while association rule learning reveals the relationships between pedestrian behaviours and infrastructure elements, highlighting the importance of facility placement and accessibility in encouraging safe crossing. PCA combined with clustering effectively reduces data complexity, revealing key factors that influence pedestrian safety. The findings emphasize the need for appropriate infrastructure, such as strategically placed zebra crossings and central refuges, to guide pedestrian behaviour and reduce accident risks. Underutilized facilities like footbridges and subways require redesign to align with pedestrian preferences. The results of this study offer insights for urban planners to prioritize safety measures and infrastructure improvements that enhance pedestrian safety at unsignalized crossings.