Geographically Weighted Comedian method for spatial outlier detection

A spatial outlier is defined as an object whose non-spatial attributes are different from the other objects in its spatial neighborhood. In this paper, a new geographically weighted method which is based on the Comedian approach for the detection of spatial outliers in a multivariate structure, called the geographically weighted comedian (GWCOM) method, is developed and discussed. A simulation study is carried out to assess and compare its performance with the existing geographically weighted methods (geographically weighted Mahalanobis distance and geographically weighted principal component analysis methods). Also, a real-life data application of the GWCOM method on a Water Quality dataset is discussed to demonstrate its effectiveness. The proposed GWCOM method shows very promising results both in simulation study as well as in application to the real data and outperforms the existing methods.