基于分解策略处理多分类不均衡问题的方法

针对多分类不均衡问题，提出了一种新的基于一对一（one-versus-one，OVO）分解策略的方法。首先基于OVO分解策略将多分类不均衡问题分解成多个二值分类问题；再利用处理不均衡二值分类问题的算法建立二值分类器；接着利用SMOTE过抽样技术处理原始数据集；然后采用基于距离相对竞争力加权方法处理冗余分类器；最后通过加权投票法获得输出结果。在KEEL不均衡数据集上的大量实验结果表明，所提算法比其他经典方法具有显著的优势。

Method based on decomposition strategy for handling multi-class imbalance problems

This paper proposed a new approach based on decomposition strategy to deal with multi-class imbalance classification problems. The method first divided the original multi-class classification problem into several binary class subproblems by using one versus scheme. Next, binary classifiers were built by employing the classification algorithms for binary class imbalance problems. Then, the SMOTE algorithm was used to deal with the original dataset and distance-based relative competence weighting method was considered to manage the non-competent classifiers. Finally, the weighted voting is employed to obtain the outputs. Experiments on several imbalanced datasets selected from the KEEL dataset repository indicate that the performance of the proposed method is better than other state-of-the-art methods.