A Preliminary Study on the Selection of Generalized Instances for Imbalanced Classification

Learning in imbalanced domains is one of the recent challenges in machine learning and data mining. In imbalanced classification, data sets present many examples from one class and few from the other class, and the latter class is the one which receives more interest from the point of view of learning. One of the most used techniques to deal with this problem consists in preprocessing the data previously to the learning process. This contribution proposes a method belonging to the family of the nested generalized exemplar that accomplishes learning by storing objects in Euclidean n-space. Classification of new data is performed by computing their distance to the nearest generalized exemplar. The method is optimized by the selection of themost suitable generalized exemplars based on evolutionary algorithms. The proposal is compared with the most representative nested generalized exemplar learning approaches and the results obtained show that our evolutionary proposal outperforms them in accuracy and requires to store a lower number of generalized examples.