Prototype Selection Using Self-Organizing-Maps and Entropy for Overlapped Classes and Imbalanced Data

The k nearest neighbor kNN is a traditional supervised classifier used in data mining tasks. However, when used in real applications, mainly in a dataset with class imbalance or class overlap, kNN suffers with problems in accuracy performance. In this paper, we propose three prototype selection methods using self-organizing maps (SOM) and information entropy to increase the effectiveness of the kNN classifier in datasets with these conditions. The methods, named SOMEntropyKnn, were able to increase the effectiveness of the kNN classifier in all the 14 datasets used in the experiment, increasing the accuracy performance from datasets with imbalance or overlap problems.