The Minkowski approach for choosing the distance metric in geographically weighted regression

In this study, the geographically weighted regression (GWR) model adapted to benefit from a broad range of distance metrics, it is demonstrated that a well-chosen distance metric can model performance. How to choose or define such a metric is key, and in this respect, a ‘Minkowski approach’ proposed that enables the selection of an optimum distance for a given GWR model. This approach is evaluated within a experiment consisting of three scenarios. The results twofold: (1) a well-chosen distance metric can significantly the predictive accuracy of a GWR model; and (2) the allows a good approximation of the underlying ‘optimal metric’, which is considered useful when the ‘true’ distance is unknown.