Imbalanced Research of Deep Belief Network Based on Dynamic Cost Sensitive

Class imbalance is a common problem in the real-world, such as fraud detection, disease diagnosis and crime rate prediction, i.e. This imbalance makes it challenging for a classifier to appropriately learn the discriminating boundaries of the majority and minority classes. To make prediction, it is necessary to solve the problem of handling imbalanced data. In this paper, we propose a cost-sensitive deep belief network, which can automatically generate and optimize the misclassification cost because differential evolution method is used. Moreover, this approach is applicable to both binary and multiclass problems. Experimental results indicate that the proposed approach can improve the performance of classification tasks in terms of two popular metrics for imbalanced classification, i.e., accuracy and F-measure.