一种基于原型学习的自适应概念漂移分类方法

发布时间：2018-04-27 09:36

本文选题：数据流 + 概念漂移　；参考：《北京邮电大学学报》2017年03期

【摘要】：为了更准确快速地处理或适应概念漂移,提出了基于原型学习的数据流分类算法,基于发掘并优化现有方法存在的问题,提出了新的方法模型Sync Prototype,在预测方法、原型判定与更新方法等处理概念漂移问题的关键部分做出了新的尝试与优化.实验结果证明,相较于现有方法,Sync Prototype模型在分类性能、概念漂移的响应速度以及时间性能等方面都有明显提高,能够更加有效处理并适应数据流概念漂移问题.
[Abstract]:In order to deal with or adapt to the concept drift more accurately and quickly, a data stream classification algorithm based on prototype learning is proposed. Based on the problems of existing methods, a new method model, Sync prototype, is proposed. New attempts and optimizations have been made to deal with the key parts of concept drift problem such as prototype decision and update method. The experimental results show that compared with the existing methods, the Sync Prototype model can improve the classification performance, the response speed of concept drift and the performance of time obviously, and can deal with and adapt to the problem of conceptual drift of data flow more effectively.
【作者单位】：北京邮电大学网络体系构建与融合北京市重点实验室;中国电力科学研究院;
【基金】：国家电网公司科技项目(XT71-15-056)
【分类号】：TP311.13

【相似文献】