Modifying the learning rate of FLNG dealing with imbalanced datasets

There are several successful approaches dealing with imbalanced datasets. In this paper, the Fuzzy Labeled Neural Gas (FLNG) is extended to work with this type of data. The proposed approach is based on assigning two different values in the learning rate depending on the data vector membership of the class. The technique is tested with several datasets and compared with other approaches. The results seem to prove that FLNG with different rates is a suitable tool for classification with a high degree of accuracy using g-means metric.