Sustainable Street Planning: A DEMATEL Based Review of Pedestrian Flow Prediction Using Machine Learning

Abstract

Rapid urbanization has intensified pedestrian congestion on commercial streets, where high footfall, mixed land use, and a limited right-of-way place mobility, safety, and commercial vitality in continual conflict. This paper reviews machine learning (ML) approaches to pedestrian flow prediction and reads them alongside a Decision-Making Trial and Evaluation Laboratory (DEMATEL) analysis of the factors that govern sustainable commercial street planning. The DEMATEL outputs reported here are conceptual and hypothesis-generating rather than empirically estimated causal weights: prominence and relation values synthesise the direction and strength of influence reported across the reviewed literature and are offered as testable propositions, not as demonstrated measurements. Ten evaluation factors, from walkability and accessibility through to smart infrastructure and real-time data availability, were drawn from the reviewed literature and sorted into provisional cause and effect groups by the DEMATEL procedure. Smart infrastructure, real-time data availability, accessibility, and land-use diversity appear to act as principal causal drivers, while pedestrian safety, walkability, social interaction, and urban livability appear to behave as dependent outcomes. On this basis the paper proposes a conceptual structure linking ML-based prediction to causal factor analysis and to planning decisions, and maps its contributions to Sustainable Development Goals (SDGs) 3, 9, 11, and 13. The review clarifies how predictive modelling and causal reasoning might be combined in support of pedestrian-centred street planning, and sets out priorities for future empirical work: structured expert elicitation, real pedestrian-flow data, and field application to case-based commercial corridors, through which the proposed cause-effect hierarchy can be tested rather than assumed.