Spatially–encouraged spectral clustering: a technique for blending map typologies and regionalization

Clustering is a central concern in geographic data science and reflects large, active domain of research. In spatial clustering, it often challenging to balance two kinds ‘goodness fit:’ clusters should have ‘feature’ homogeneity, that they aim represent one ‘type’ observation, also ‘geographic’ coherence, some detected geographical ‘place’. This divides ‘map typologization’ studies, common geodemographics, from ‘regionalization’ optimization statistics. Recent attempts simultaneously typologize regionalize into with both feature homogeneity coherence faced conceptual computational challenges. Fortunately, new work on spectral clustering can address regionalization typologization tasks within the same framework. research develops novel kernel combination method for use allows analysts blend smoothly between coherence. I explore formal properties methods recommend multiplicative clustering. Altogether, spatially encouraged shown as order reveal geographies latent structured data.