中文摘要：

分类预测是数据挖掘和机器学习的重要任务之一，非均衡数据广泛存在于真实世界的分类问题中．本文提出一种新的解决非均衡数据集的预处理方法（ImSMOTE-RSTR.）．通过改进的SMOTE方法创建新的人工合成少数类实例，并在此基础上应用基于粗糙集理论的子集下近似技术对训练集进行清理，该算法被验证得到较理想的结果．

英文摘要：

Classification forecasting is an important work of data mining and machine learning, imbalanced data is a common problem in classification in most real domains. This paper proposes a new hybrid method for preprocessing imbalanced datasets （ImSMOTE-RSTR.） through the construction of new samples, using the improved Synthetic Minority Oversampling Technique together with the application of an editing technique based on the Rough Set Theory and the lower approximation of a subset. The proposed method has been validated by an experimental study showing good results as the learning algorithm.