Identifying street-character-weighted local area using locally weighted community detection methods the case study of London and Amsterdam
Paper i proceeding, 2019
Previous research suggests that community detection methods, which defines subgraph that maximises internal ties and minimise external ties, can be applied on the street network dual graph in identifying Street-based Local Area (Law et al 2016; Law 2017). The method was successful in identifying isolated local area but were unsuccessful in identifying local area that was less driven by the grid but more from other urban factors. This research attempts to address this problem by embedding street characteristics in community detection to define Street Character Weighted Local Area (St-W-LA). The idea is that street neighbourhoods are not only defined by the topology of the street network but also by the morphology of the built form. In particular, we adopted Spacemate Building Density Metric in defining Density-based local area for Amsterdam in the Netherlands and Space Syntax angular choice metric in defining angular-choice-based (note. for simplicity reasons we term this betweeness-based) local area for London in the UK. We compared the results of the community detection with user defined local area through visual analysis. In general, we found the weighted and the unweighted street-based local areas to be similar. This suggests that neighbourhood characteristics (morphology) follow neighbourhood topology where areas that were built in similar times with similar density and building type were also better connected internally. However, we also found notable differences between the two methods where the weighted local area seems better in capturing the user defined local area in more continuous grids. Further empirical research employing mental map studies and intra-cluster analysis are needed to validate the method.