A Novel Technique on Class Imbalance Big Data using Analogous under Sampling Approach

Abstract

In this paper, we propose hybrid Random under Sampled Imbalance Big Data (USIBD) framework to extract knowledge from class imbalance big data.A novel undersampling method for the base learner is also proposed to handle the dynamic class-imbalance problem caused by the gradual evolution of classes in big data.The proposed USIBD knowledge discovery framework is robust and less sensitive to outliers where non-uniform distribution of data is applied.Empirical studies demonstrate the effectiveness of USIBD in various class imbalance big datasets scenarios in comparison to existing methods.