Spatial clustering with different geographical scales
Alberto A.E. Jakob, Universidade Estadual de Campinas (UNICAMP)
This paper aims to provide a clustering methodology which uses not only tabular information, but spatial information also in order to create more homogeneous areas in certain space. The local Moran’s I can be used as a spatial autocorrelation indicator, and gives information about the spatial unit also in terms of its neighbors. Spatial clustering techniques are used to different geographical scales to define homogeneous areas according to certain characteristics. The results for both scales are correlated and adjusted if necessary. The idea is to adjust boundaries created by less detailed geographical areas (with more tabular data) in terms of those created by census tracts, with more detail but less tabular data. The result is a map with accurate boundaries and available to use with all demographic census data.