基于Boosting的迭代加权集成分类算法

Boosting-based iterative weighted ensemble classification algorithm

在集成分类中,如何对基分类器实现动态更新和为基分类器分配合适的权值一直是研究的重点。针对以上两点,提出了BIE和BIWE算法。BIE算法通过最新训练的基分类器的准确率确定集成是否需要替换性能较差的基分类器及需替换的个数,实现对集成分类器的动态迭代更新;BIWE算法在此基础上提出了一个加权函数,对具有不同参数特征的数据流可以有针对性地获得基分类器的最佳权值,从而提升集成分类器的整体性能。实验结果表明,BIE算法相较对比算法在准确率持平或略高的情况下,可以减少生成树的叶子数、节点数和树的深度;BIWE算法相较对比算法不仅准确率较高,而且能大幅度减少生成树的规模。

In ensemble classification,how to dynamically update the base classifier and assign appropriate weights to the base classifier has always been the focus of research.In view of the above two points,this paper proposed BIE and BIWE algorithm.Based on the accuracy of the latest trained base classifiers,the BIE algorithm determined whether or not to replace the base classifiers with poor performance and the number of them,so as to realize the dynamic iterative update of the ensemble classifier.Based on BIE,BIWE algorithm proposed a weighting function,which could specifically obtain the optimal weight of the base classifier for data streams with different parameter characteristics,in order to improve the overall performance of the ensemble classifier.Experimental results show that the BIE algorithm can reduce the number of leaves,nodes and depth of the spanning tree when the accuracy is the same or slightly higher;the BIWE algorithm not only has higher accuracy,but also can greatly reduce the size of the spanning tree.