Clustering with implicit constraints: A novel approach to housing market segmentation

Constrained clustering has been widely studied and outperforms both the traditional unsupervised clustering and experience-oriented approaches. However, the existing literature on constrained clustering concentrates on spatially explicit constraints, while many constraints in housing market studies are implicit. Ignoring the implicit constraints will result in unreliable clustering results. This article develops a novel framework for constrained clustering, which takes implicit constraints into account. Specifically, the research extends the classical greedy searching algorithm by adding one back-and-forth searching step, efficiently coping with the order sensitivity. Via evaluation on both synthetic and real data sets, it turns out that the proposed algorithm outperforms existing algorithms, even when only the traditional pairwise constraints are provided. In an application to a concrete housing market segmentation problem, the proposed algorithm shows its power to accommodate user-specified homogeneity criteria to extract hidden information on the underlying urban spatial structure.