In this paper, we propose a novel adaptive learning algorithm
ADASYN for imbalanced data classification problems.
Based on the original data distribution, ADASYN can
adaptively generate synthetic data samples for the minority
class to reduce the bias introduced by the imbalanced data
distribution. Further more, ADASYN can also autonomously
shift the classifier decision boundary to be more focused on
those difficult to learn examples, therefore improving learning
performance. These two objectives are accomplished by a
dynamic adjustment of weights and an adaptive learning
procedure according to data distributions. Simulation results
on five data sets based on various evaluation metrics show the
effectiveness of this method.
Imbalanced learning is a challenging and active research
topic in the artificial intelligence, machine learning, data
mining and many related areas. We are currently investigating
various issues, such as multiple classes imbalanced learning
and incremental imbalanced learning. Motivated by the results
in this paper, we believe that ADASYN may provide a
powerful method in this domain.