DBIG-US: A two-stage under-sampling algorithm to face the class imbalance problem

The class imbalance problem occurs when one far outnumbers the other classes, causing most traditional classifiers perform poorly on minority classes. To tackle this problem, a plethora of techniques have been proposed, especially centered around resampling methods. This paper introduces two-stage method that combines DBSCAN clustering algorithm to filter noisy majority instances with graph-based procedure overcome imbalance. We then experimentally evaluate behavior proposed collection two-class imbalanced data sets. experimental results show an improvement in classification performance measured by geometric mean accuracy each and also higher reduction ratio compared several state-of-the-art under-sampling techniques.