Generally, class imbalance is a major problem that occurs with various machine learning problems where there are a large number of instances of one class compared to another class. In this case, it has been found that a typical classification model may be biased towards a particular class and may not be able to classify instances of another class properly. This problem is critical while dealing with various real-world problems. In recent times, Broad Learning System (BLS) has gained attention from many researchers due to its fast training speed and efficiency of the flat network. However, it has been found that the conventional BLS model may not be efficient while dealing with imbalanced data sets because it uses a least squares method. Generally, a least squares method may not be efficient while dealing with imbalanced data sets.

This paper proposes an adaptive margin broad learning system (AMBLS) model to improve its efficiency while handling imbalanced data sets in order to address this issue with traditional BLS models. The suggested approach may be more effective when handling imbalanced data sets than traditional BLS models, according to experimental results on a number of benchmark imbalanced data sets.

  Keywords- Imbalanced Learning; Broad Learning System; Adaptive Margin; Minority Classification; Machine Learning.