2018-12-172018-12-172017-03-24LIMA, Andrey Marcos Souza da Silva de. Improving diffraction pattern recognition using kNN. Orientador: Daniel Leal Macedo. 2017. 46 f. Trabalho de Curso (Bacharelado em Geofísica) - Faculdade de Geofísica, Instituto de Geociências, Universidade Federal do Pará, Belém, 2017. Disponível em: <http://bdm.ufpa.br/jspui/handle/prefix/893>. Acesso em:.http://bdm.ufpa.br/jspui/handle/prefix/893This work reinforces the importance of using pattern recognition in order to classify seismic events such as diffractions, edge diffractions, reflections and void points. Their identification and processing can be used for the construction of velocity models and the imaging of geological structures. The diffraction operator responses are analyzed using an algorithm to calculate two pairs of three parameters that characterize an event. The k nearest neighbor method (kNN) is used to classify these events as diffractions, reflections, edge diffractions and void points based on their diffractor operator. Since the kNN method uses a measure of distance, this work compares the classification using Euclidean and Mahalanobis distances. The results showed that e10-e6-e3 domain using Mahalanobis distance is the best combination to better cluster and classify events.Acesso AbertoProspecção SísmicaDifraçãoGeofísicaCNPQ::CIENCIAS EXATAS E DA TERRA::GEOCIENCIAS::GEOFISICA::SISMOLOGIATrabalho de Curso - Graduação - Monografia