期刊名称:Journal of Theoretical and Applied Information Technology
印刷版ISSN:1992-8645
电子版ISSN:1817-3195
出版年度:2018
卷号:96
期号:9
出版社:Journal of Theoretical and Applied
摘要:Concept drift is a known problem that can occur in classifier systems. Detecting and handling concept drift is an active area of research. Once a concept drift is detected, it has to be handled by updating or re-generating the classification model. In this paper, a new approach is introduced for handling concept drift, where a drift intensity measure is used to quantify the intensity of a concept drift. The model generation process uses the drift intensity measure while generating a new model. If the drift intensity is high, the model generation process discards old data (data before the drift occurrence) and builds a new model solely based on the new data after drift. On the other hand, if the drift intensity is low or moderate, the model generation process takes into account both old data and new data but it gives more weight (proportional to the drift intensity) to the new data as compared to old data.
关键词:Data Mining; Classification; Concept Drift; Drift Handling; Big Data.