基于随机平衡采样的不平衡数据流分类研究

数据流广泛应用于现实世界的多个领域,但是不平衡数据流的存在严重影响了传统数据流分类器的性能.针对不平衡数据流问题,提出了随机平衡采样算法(RBS)处理数据流的不平衡问题,并以RBS算法为基础提出了随机平衡采样数据流集成算法(RBSSEA)旨在解决不平衡数据流的分类问题.最后,分别采用合成和真实数据集对RBSSEA算法进行验证,实验结果证明RBSSEA算法在解决不平衡数据流分类问题具有一定的优势.

A study of the classification of imbalanced data streams based on random balance sampling

" Data stream" proposed by Henzinger is now widely used in the real world. However,imbalanced data streams seriously affect the functions of the traditional datastream classifiers. This paper proposes a new algorithm based on random balance sampling(RBS) to solve this problem. It then proposes an ensemble algorithm based on the random balance sampling of data streams for the problem of classifiers. The experiments based on synthetic and real-world datasets show that this new ensemble algorithm is fairly efficient for solving the problem of imbalanced data streams.