基于信息熵的概念漂移检测

Concept drift detection based on information entropy

文章针对概念漂移检测分类器很难维持较高的分类性能,存在错误检测和延迟检测等问题,提出了一种基于信息熵的概念漂移检测算法。首先,使用信息熵对动态数据流中的概念漂移进行检测;然后,将检测到的概念漂移信息,在概念池中进行汇总和统计;最后,使用了两种公开的真实数据和一种人造概念漂移数据进行实验,并对实验结果进行分析,验证了模型的有效性和正确性。实验结果表明,该算法可以有效地检测概念漂移和更新分类器,同时表现出较好的分类性能。

Due to problems such as error detection and delay detection,the classifier of concept drift detection can not maintain a higher classification performance.In this study,a concept drift detection algorithm based on information entropy was proposed.Firstly,the concept drift of dynamic data stream is detected with information entropy.Secondly,the detected concept drift information will be collected and counted in the concept pool.Finally,two publicly available real data and an artificial conceptual drift data are used to make experiments and the experimental results are analyzed to verify the validity and correctness of the model.The results indicate that the proposed algorithm can effectively detect concept drifts and update classifiers,which show a good classification performance.