In a point process spatio-temporal framework, we consider the problemof features detection in the presence of clutters. We extend the methodology of Byersand Raftery (1998) to the spatio-temporal context by considering the propertiesof the K-th nearest-neighbour distances. We make use of the spatio-temporal distancebased on the Euclidean norm where the temporal term is properly weighted.We show the form of the probability distributions of such K-th nearest-neighbourdistance. A mixture distribution, whose parameters are estimated with an EM algorithm,is used to classify points into clutters or features. We assess the performanceof the proposed approach with a simulation study, together with an application toearthquakes.
|Title of host publication||Smart Statistics for Smart Applications Book of short paper|
|Number of pages||8|
|Publication status||Published - 2019|