A novel soft spatial weights matrix method based on soft sets

Soft sets are efficient and flexible tools to describe uncertainty and fuzziness. In this paper, we integrate soft set theory into the specification of fuzzy spatial dependent relationship and provide a framework for spatial weight description. We treat the configuration of spatial location relationship as soft sets and propose a new dependence measure based on operations of soft sets. The proposed soft spatial weights matrix efficiently combines information of 'spatial adjacent relation' and 'spatial distance'. Further, Chinese regional industrial agglomeration data are applied to empirical analysis. The spatial autoregressive error panel model (SEM) with our new matrix performs better than that of Moran's I, log-likelihood and interpretation.