A density weighted fuzzy outlier clustering approach for class imbalanced learning

Abstract

The class imbalance problem is widely studied in the machine learning community, and it is present in many real-world applications such as spam filtering, anomaly detection and medical diagnosis. In this paper, we propose a density weighted fuzzy outlier clustering approach for class imbalanced learning. The method considers a novel fuzzy neighborhood relation with local density information when assigning the weights to the samples in the clustering process, and it is then hybridized with the fuzzy outlier clustering approach for a novel fuzzy clustering method. In this way, the most representative majority class samples are chosen while the outlier samples are subjected to elimination. The validity of the proposed method is tested with synthetic and real-world datasets which demonstrates superior performance compared to other clustering-based resampling schemes. Thus, the density weighted fuzzy outlier clustering approach can be used for real life imbalanced problems.