Exploring the Associations of Socioeconomic Characteristics and Distance Decay Effects in Spatial Interaction

Abstract

The Spatial Interaction (SI) model is a prominent tool for predicting trip flows based on distance decay. Despite extensive discussions on spatial heterogeneity and spatial structure, existing SI models are still exploring ways to incorporate local distance decay variations within small urban areas. Furthermore, non-spatial factors, such as socioeconomic characteristics, are typically underestimated in SI and other travel flow prediction models. To tackle these issues, this study introduces a novel two-step SI model that enhances travel flow predictions. It first utilises a k-means clustering algorithm to group areas based on residents' socioeconomic characteristics, then calibrates the localised distance-decay parameter in the origin-specific gravity model for each group and predicts the travel flows. Demonstrated by a case study of the Greater London Area (GLA), we uncovered local distance decay patterns in commuting trips and explained their associations with spatial structure and non-spatial factors using census data. Most importantly, the results proved that our two-step SI model could significantly improve the accuracy of flow predictions without considerably increasing computational complexity.

Keywords: Distance decay effects, spatial interaction model, socioeconomic characteristic, commuting behaviour, clustering algorithm.