Geographically Weighted Non-negative Principal Components Analysis for Exploring Spatial Variation in Multidimensional Composite Index. GeoComputation 2019
conference contributionposted on 01.12.2019 by Narumasa Tsutsumida, Daisuke Murakami, Takahiro Yoshida, Tomoki Nakaya, Binbin Lu, Paul Harris
Any type of content contributed to an academic conference, such as papers, presentations, lectures or proceedings.
The objectives of this study are to develop geographically weighted (GW) non-negative principal components analysis (PCA) and to explore spatial variations of contributions to a multidimensional composite index (MCI) from spatial multidimensional data. As a case study, we produced a MCI for earthquake risk in Tokyo, Japan, in 2018 from the collapse risk of buildings, the fire risk, and the evacuation risk associated with earthquake. GW non-negative PCA was applied to these data to uncover spatial variation of non-negative weightings (eigenvectors). This study demonstrates GW non-negative PCA provides more informative outputs when considering local differences of contributions of multidimensional data to the composite index.