The SUPRAIC Algorithm: A Suppression Immune Based Mechanism to Find a Representative Training Set in Data Classification Tasks

This article proposes a new classifier inspired on a biological immune systems’ characteristic. This immune based predictor also belongs to the class of k-nearest-neighbors algorithms. Nevertheless, its main features, compared to other artificial immune classifiers, are the assumption that training set is the antibodies’ population and a suppression mechanism that tries to reduce the training set into a smaller subset. This subset is supposed to contain the most significative samples, without loosing much capability of generalization. It is known that in prediction problems, the choice of a good training set is crucial for the classification process. And this is the focus of this research. Experiments using some benchmarks and the analysis of the results of our ongoing work are presented.