AGURF: An adaptive general unified representation frame for imbalanced interval-valued data

Abstract

Interval-valued data (IVD) is a kind of data in which each feature is an interval, and embeds some uncertainty and variability information. Due to the inherent structural particularity of IVD, in addition to the skewed distribution classes, another imbalance for IVD is the biased range distribution on each interval-valued data. This internal imbalance depicts the internal distribution of IVD in detail, but is usually ignored in representation. This work proposes an adaptive general unified representation frame (AGURF), which may expand the representation frame of IVD. Based on the unified representation frame (URF) proposed in previous work, an adaptive general unified representation frame is constructed firstly. Then the offset-center is defined to re-measure the location of each interval-valued data more accurate. Meanwhile, a rule to set the adaptive factors for each class automatically, which serves as base factors to balance the relationship between offset-center and radius, is proposed. Finally, several general classifiers are also used to verify AGURF. The experiment results on synthetic and real-world datasets demonstrate that the proposed method can better represent imbalanced IVD, obtain good classification performance and reduce time cost simultaneously.