Merge-Based Prototype Selection for Nearest Neighbor Classification

A generalized prototype-based classification scheme founded on hierarchical clustering is proposed. The basic idea is to obtain a condensed 1-NN classification rule by replacing a group of prototypes by a representative while approximately keeping its original classification abilities. The algorithm improves and generalizes previous works by explicitly introducing the concept of cluster and cluster consistency. Apart from the quality of the obtained sets, the proposed scheme permits a very efficient and flexible implementation by using geometric cluster properties and different intercluster measures. Moreover, the algorithm benefits from all the well-known results about hierarchical clustering regarding computational improvements and different intercluster measures. Empirical results demonstrate the merits of the proposed algorithm taking into account the size of the condensed sets of prototypes, the accuracy of the corresponding condensed 1-NN classification rule and the computing time.