Learning efficient and interpretable prototypes from data for nearest neighbor classification method

The Nearest Neighbor Classification (NNC) has been widely used as classification method, due to its simplicity, classification efficiency and its ability to deal with different classification problems. Despite its good classification accuracy, the NNC suffers from many shortcomings on the execution time, noise sensitivity, high storage requirements and lack of interpretability. In this paper, we propose a new prototype learning algorithm for NNC method. This new algorithm derives a small subset of relevant prototypes from the training set by the means of clustering algorithms. The obtained prototypes enable the NNC method to deal with the aforementioned issues. Moreover, using clustering algorithm involve an interpretable set of prototypes that can describes the application domain. The proposal is experimented on various datasets and compared to other prototypes learning methods. The results show the effectiveness of our approach compared to the existed methods.