Multilabel Prototype Generation for data reduction in K-Nearest Neighbour classification

• Multilabel Prototype Generation for efficient k-Nearest Neighbour classification Four multiclass methods are adapted to the multilabel space Evaluation with twelve corpora, three noise scenarios, and different classifiers Proposals significantly improve performance reduction rates of reference strategies Novel adaptations proposed also show significant capabilities (PG) typically considered improving efficiency k -Nearest ( NN) classifier when tackling high-size corpora. Such approaches aim at generating a reduced version corpus without decreasing compared initial set. Despite their large application in very few works have addressed proposal PG space. In this regard, work presents novel adaptation four case. These proposals evaluated NN-based classifiers, 12 corpora comprising varied range domains sizes, scenarios artificially induced data. The results obtained that capable improving—both terms performance—the only literature as well case which no method is applied, presenting statistically superior robustness noisy scenarios. Moreover, these allow prioritising either or efficacy criteria through its configuration depending on target scenario, hence covering wide area solution not previously filled by other works.