In this paper, a set of neighbourhood-based classifiers are jointly used in order to select a more reliable neighbourhood of a given sample and take an appropriate decision about its class membership. The approaches introduced here make use of two concepts: proximity and symmetric placement of the samples.