Adaptive Subspace Optimization Ensemble Method for High-Dimensional Imbalanced Data Classification

It is hard to construct an optimal classifier for high-dimensional imbalanced data, on which the performance of classifiers is seriously affected and becomes poor. Although many approaches, such as resampling, cost-sensitive, and ensemble learning methods, have been proposed to deal with the skewed data, they are constrained by high-dimensional data with noise and redundancy. In this study, we propose an adaptive subspace optimization ensemble method (ASOEM) for high-dimensional imbalanced data classification to overcome the above limitations. To construct accurate and diverse base classifiers, a novel adaptive subspace optimization (ASO) method based on adaptive subspace generation (ASG) process and rotated subspace optimization (RSO) process is designed to generate multiple robust and discriminative subspaces. Then a resampling scheme is applied on the optimized subspace to build a class-balanced data for each base classifier. To verify the effectiveness, our ASOEM is implemented based on different resampling strategies on 24 real-world high-dimensional imbalanced datasets. Experimental results demonstrate that our proposed methods outperform other mainstream imbalance learning approaches and classifier ensemble methods.

I. Introduction

The problem of class-imbalanced data is often encountered in many fields of machine learning and data mining, such as bioinformatics mining [1], [2], gene data analysis [3], [4], disease diagnosis [5], image recognition [6], [7], and text classification [8], [9]. For binary imbalanced data, the number of majority-class samples is significantly larger than that of the minority-class samples, which may hurt traditional classification algorithms [10]. Because standard classifiers focus on maximizing the overall classification accuracy, they are biased toward the majority class for the skewed data [11], [12]. As a result, it is difficult to identify samples of the minority class correctly. For example, it can obtain an overall classification accuracy of 99% when a classification algorithm predicts all samples as the majority class, where 1% of samples are belonging to the minority class. Unfortunately, all the samples belonging to the minority class are misclassified. In this setting, the minority-class samples are more worthy of attention [13]–[16], because the misclassification of them often leads to more serious consequences. Similar to most imbalance learning studies, our method focuses on the binary class-imbalanced problem. Furthermore, for high-dimensional imbalanced data, a large number of noisy and redundant features make the classification algorithm suffer from greater challenges [17], [18]. Therefore, it is necessary to devise an effective approach to deal with high-dimensional imbalanced data.