A Kernel Clustering-Based Possibilistic Fuzzy Extreme Learning Machine for Class Imbalance Learning

Abstract

Compared with traditional computational intelligence techniques such as the support vector machine, extreme learning machine (ELM) provides better generalization performance at a much faster learning speed without tuning model parameters. Unfortunately, the training process of ELM is still sensitive to the outliers or noises in the training set. On the other hand, when it comes to imbalanced datasets, ELM produces suboptimal classification models. In this paper, a kernel possibilistic fuzzy c-means clustering-based ELM algorithm for class imbalance learning (CIL) is developed to handle the class imbalance problem in the presence of outliers and noises. A set of experiments are conducted on several artificial and real-world imbalanced datasets for testing the generalization performance of the proposed algorithm. Additionally, we compare its performance with some typical CIL methods. The results indicate that the proposed method is a very effective method for CIL, especially in the presence of outliers and noises in datasets.