³ÉÈË´óƬ

This talk focused on one of the most challenging issues in non-stationary data streams, known as concept drift. Concept drift occurs when the pattern of data changes over time, making models trained on historical data less effective at predicting new data patterns. This inability to adapt compromises the performance of well-trained machine learning models. From the perspective of concept drift adaptation, this presentation will introduce three general strategies for updating learned models with newly arrived data, aimed at mitigating the impact of concept drift.