Classification of imbalanced data sets by using the hybrid re-sampling algorithm based on isomap

Abstract

The majority of machine learning algorithms previously designed usually assume that their training sets are well-balanced, but data in the real-world is usually imbalanced. The class imbalance problem is pervasive and ubiquitous, causing trouble to a large segment of the data mining community. As the conventional machine learning algorithms have bad performance when they learn from imbalanced data sets, it is necessary to find solutions to machine learning on imbalanced data sets. This paper presents a novel Isomap-based hybrid re-sampling approach to improve the conventional SMOTE algorithm by incorporating the Isometric feature mapping algorithm (Isomap). Experiment results demonstrate that this hybrid re-sampling algorithm attains a performance superior to that of the re-sampling. It is clear that the Isomap method is an effective means to reduce the dimension of the re-sampling. This provides a new possible solution for dealing with the IDS classification. © Springer-Verlag 2009.