AUCBoost算法处理不平衡分类问题

AUC-Based boosting method for the classification of rare classes

在现实生活中很多应用都包含了对不平衡数据集的分类.由于不平衡数据集中多数类与稀有类的数量相差较大,所以大多数分类算法都不能够很好地对稀有类样本进行分类,而通常稀有类才是我们首要关心的,这就给不平衡数据的分类提出了挑战,为了更好地处理不平衡数据集的分类问题,本文提出了一种以基分类器的ROC曲线下面积(AUC面积)为分类权重的AUCBoost分类算法.

Classifying to unbalance data sets has many applications in our life.The number of rare class is much less than the other,so most classiers could not do well when they meet rare class.But usually the recognition of the rare class is most important,so it's really a challenge for classification problem of unbalance data sets.To solve the problem better,the paper proposes a novel AUC-Based(AUC: Area Under ROC Curve) Boosting method.